# Appendix Source: https://www.recraft.ai/docs/api-reference/appendix ### Maximum prompt length The maximum prompt length depends on the model. | Model | Limit | | --------------------- | ----- | | Recraft V4 | 10000 | | Recraft V4 Vector | 10000 | | Recraft V4 Pro | 10000 | | Recraft V4 Pro Vector | 10000 | | Recraft V3 | 1000 | | Recraft V3 Vector | 1000 | | Recraft V2 | 1000 | | Recraft V2 Vector | 1000 | ### List of supported image sizes The image size can be specified using one of two formats: * Explicit dimensions (`WxH`): defines the exact width (`W`) and height (`H`) of the output image in pixels. Example: `1820x1024`. * Aspect ratio (`w:h`): defines the proportional relationship between width and height without specifying absolute dimensions. Example: `16:9`. See [Wikipedia](https://en.wikipedia.org/wiki/Aspect_ratio_\(image\)) for more details. The set of valid aspects and size values depends on the selected model. Refer to the table below for supported configurations.
Models Aspect Size
Recraft V4 `1:1``1024x1024`
`2:1``1536x768`
`1:2``768x1536`
`3:2``1280x832`
`2:3``832x1280`
`4:3``1216x896`
`3:4``896x1216`
`5:4``1152x896`
`4:5``896x1152`
`6:10``832x1344`
`14:10``1280x896`
`10:14``896x1280`
`16:9``1344x768`
`9:16``768x1344`
Recraft V4 Pro `1:1``2048x2048`
`2:1``3072x1536`
`1:2``1536x3072`
`3:2``2560x1664`
`2:3``1664x2560`
`4:3``2432x1792`
`3:4``1792x2432`
`5:4``2304x1792`
`4:5``1792x2304`
`6:10``1664x2688`
`14:10``2560x1792`
`10:14``1792x2560`
`16:9``2688x1536`
`9:16``1536x2688`
Recraft V2
Recraft V3		 `1:1``1024x1024`
`2:1``2048x1024`
`1:2``1024x2048`
`3:2``1536x1024`
`2:3``1024x1536`
`4:3``1365x1024`
`3:4``1024x1365`
`5:4``1280x1024`
`4:5``1024x1280`
`6:10``1024x1707`
`14:10``1434x1024`
`10:14``1024x1434`
`16:9``1820x1024`
`9:16``1024x1820`
Recraft V4 Pro Vector
Recraft V4 Vector
Recraft V3 Vector
Recraft V2 Vector		 `1:1`
`2:1`
`1:2`
`3:2`
`2:3`
`4:3`
`3:4`
`5:4`
`4:5`
`6:10`
`14:10`
`10:14`
`16:9`
`9:16`
### Policies * All generated images are currently stored for approx. 24 hours, this policy may change in the future, and you should not rely on it remaining constant. * Images are publicly accessible via direct links without authentication. However, since the URLs include unique image identifiers and are cryptographically signed, restoring lost links is nearly impossible. * Currently, image generation rates are defined on a per-user basis and set at **100 images per minute**. In addition, requests are limited to **5 per second**. These rate limits may be adjusted in the future. Need help or have suggestions for improving our docs? [Contact support](mailto:\[help@recraft.ai]\(mailto:help@recraft.ai\)) # Endpoints Source: https://www.recraft.ai/docs/api-reference/endpoints Dig into the details of the Recraft API endpoints. ## Authentication We use [Bearer](https://swagger.io/docs/specification/authentication/bearer-authentication/) API tokens for authentication. To access your API key, log in to Recraft and [enter your profile](https://app.recraft.ai/profile/api). All requests should include your API key in an Authorization HTTP header as follows:\ \ `Authorization: Bearer RECRAFT_API_TOKEN` ### OpenAI Python library Future examples will be shown using OpenAI Python library, for example, once installed, you can use the following code to be authenticated: ```python theme={null} from openai import OpenAI client = OpenAI( base_url='https://external.api.recraft.ai/v1', api_key=, ) ``` Since the Recraft API follows REST principles, you can interact with it using any standard HTTP client such as `curl`, as well as your preferred programming language or library. ## Generate image Creates an image given a prompt. ```javascript theme={null} POST https://external.api.recraft.ai/v1/images/generations ``` ### Example ```python theme={null} response = client.images.generate( prompt='race car on a track', ) print(response.data[0].url) ``` ### Parameters | Parameter | Type | Description | Compatibility | | ----------------- | -------------------------------------- | ------------------------------------------------------------------------------------------ | ---------------------------------------------------------------------------------------------------------------------------------------------------------- | | prompt (required) | string | A text description of the desired image(s). | Maximum prompt length depends on model, refer to [Appendix](/api-reference/appendix#maximum-prompt-length) | | n | integer or null, default is 1 | The number of images to generate, must be between 1 and 6. | All models | | model | string or null, default is `recraftv4` | The model to use for image generation. | Must be one of: `recraftv4`, `recraftv4_vector`, `recraftv4_pro`, `recraftv4_pro_vector`, `recraftv3`, `recraftv3_vector`, `recraftv2`, `recraftv2_vector` | | style | string or null | The style of the generated images, refer to [Styles](/api-reference/styles). | V2 / V3 styles | | style\_id | UUID or null | Use a style as visual reference, refer to [Styles](/api-reference/styles). | V2 / V3 styles | | size | string or null, default is `1:1` | The size of the generated images in `WxH` or `w:h` format. | Depends on model, supported values are published in [Appendix](/api-reference/appendix#list-of-supported-image-sizes) | | negative\_prompt | string or null | A text description of undesired elements on an image. | V2 / V3 models | | response\_format | string or null, default is `url` | The format in which the generated images are returned. Must be one of `url` or `b64_json`. | All models | | text\_layout | Array of objects or null | Refer to [Text Layout](#text-layout). | V3 models only | | controls | object or null | A set of custom parameters to tweak generation process, refer to [Controls](#controls). | All models, partially supported for V4 | **Hint:** if OpenAI Python Library is used, non-standard parameters can be passed using the `extra_body` argument. For example: ```python theme={null} response = client.images.generate( prompt='race car on a track', extra_body={ 'style_id': style_id, 'controls': { ... } } ) print(response.data[0].url) ``` ## Create style Upload a set of images to create a style reference. ```javascript theme={null} POST https://external.api.recraft.ai/v1/styles ``` ### Example ```python theme={null} response = client.post( path='/styles', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, body={'style': 'digital_illustration'}, files={'file1': open('image.png', 'rb')}, ) print(response['id']) ``` ### Output ```javascript theme={null} {"id": "229b2a75-05e4-4580-85f9-b47ee521a00d"} ``` ### Request body Upload a set of images to create a style reference. | Parameter | Type | Description | | ---------------- | ------ | ------------------------------------------------------------------------------------------------------------------------------------------- | | style (required) | string | The base style of the generated images, should be one of: `any`, `realistic_image`, `digital_illustration`, `vector_illustration`, `icon` | | files (required) | files | Images in PNG, JPG, or WEBP format for use as style references. The max number of images is 5. Total size of all images is limited to 5 MB. | ## Image to image Image-to-image operation tool allows you to create images similar to a given image, preserving certain aspects like composition, color, or subject identity while altering others based on the prompt. This can be used to create variations of the image or images that have similar composition to existing image. Use prompt to describe the new image and strength to define similarity level. ***Note***: This operation is available with Recraft V3, Recraft V3 Vector models only. ```javascript theme={null} POST https://external.api.recraft.ai/v1/images/imageToImage ``` ### Example ```python theme={null} response = client.post( path='/images/imageToImage', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, files={ 'image': open('image.png', 'rb'), }, body={ 'prompt': 'winter', 'strength': 0.2, }, ) print(response['data'][0]['url']) ``` ### Parameters | Parameter | Type | Description | Compatibility | | ------------------- | -------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------ | | image (required) | file | An image to modify, must be less than 5 MB in size, have resolution less than 16 MP, and max dimension less than 4096 pixels. | | | prompt (required) | string | A text description of areas to change. | refer to [Appendix](/api-reference/appendix#maximum-prompt-length) | | strength (required) | float | Defines the difference with the original image, should lie in `[0, 1]`, where `0` means almost identical, and `1` means minimal similarity. | | | n | integer or null, default is 1 | The number of images to generate, must be between 1 and 6. | | | model | string or null, default is `recraftv3` | The model to use for image generation. | Must be one of: `recraftv3`, `recraftv3_vector` | | style | string or null | The style of the generated images, refer to [Styles](/api-reference/styles). | V3 styles only | | style\_id | UUID or null | Use a style as visual reference, refer to [Styles](/api-reference/styles). | V3 styles only | | response\_format | string or null, default is `url` | The format in which the generated images are returned. Must be one of `url` or `b64_json`. | | | negative\_prompt | string or null | A text description of undesired elements on an image. | | | text\_layout | Array of objects or null | Refer to [Text Layout](#text-layout). | | | controls | object or null | A set of custom parameters to tweak generation process, refer to [Controls](#controls). | | ## Image inpainting Inpainting lets you regenerate specific parts of an image while keeping the rest intact. This is useful for correcting details, replacing elements, or making creative changes without starting from scratch. It uses a mask to identify the areas to be filled in, where white pixels represent the regions to inpaint, and black pixels indicate the areas to keep intact, i.e. the white pixels are filled based on the input provided in the prompt. ***Note***: This operation is available with Recraft V3, Recraft V3 Vector models only. ```javascript theme={null} POST https://external.api.recraft.ai/v1/images/inpaint ``` ### Example ```python theme={null} response = client.post( path='/images/inpaint', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, files={ 'image': open('image.png', 'rb'), 'mask': open('mask.png', 'rb'), }, body={ 'prompt': 'winter', }, ) print(response['data'][0]['url']) ``` ### Parameters Body of a request should contain an image file and a mask in PNG, JPG or WEBP format and parameters passed as content type `'multipart/form-data'`. The image must be no more than 5 MB in size, have resolution no more than 16 MP, max dimension no more than 4096 pixels and min dimension no less than 256 pixels. | Parameter | Type | Description | Compatibility | | ----------------- | -------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------ | | image (required) | file | An image to modify, must be less than 5 MB in size, have resolution less than 16 MP, and max dimension less than 4096 pixels. | | | mask (required) | file | An image encoded in **grayscale color mode**, used to define the specific regions of an image that need modification. The white pixels represent the parts of the image that will be inpainted, while black pixels indicate the parts of the image that will remain unchanged. Should have exactly the same size as the image. Each pixel of the image should be either pure black (value `0`) or pure white (value `255`). | | | prompt (required) | string | A text description of areas to change. | refer to [Appendix](/api-reference/appendix#maximum-prompt-length) | | n | integer or null, default is 1 | The number of images to generate, must be between 1 and 6. | | | model | string or null, default is `recraftv3` | The model to use for image generation. | Must be one of: `recraftv3`, `recraftv3_vector` | | style | string or null | The style of the generated images, refer to [Styles](/api-reference/styles). | V3 styles only | | style\_id | UUID or null | Use a style as visual reference, refer to [Styles](/api-reference/styles). | V3 styles only | | response\_format | string or null, default is `url` | The format in which the generated images are returned. Must be one of `url` or `b64_json`. | | | negative\_prompt | string or null | A text description of undesired elements on an image. | | | text\_layout | Array of objects or null | Refer to [Text Layout](#text-layout). | | | controls | object or null | A set of custom parameters to tweak generation process, refer to [Controls](#controls). | | ## Replace background Replace Background operation detects background of an image and modifies it according to given prompt. ***Note***: This operation is available with Recraft V3, Recraft V3 Vector models only. ```javascript theme={null} POST https://external.api.recraft.ai/v1/images/replaceBackground ``` ### Example ```python theme={null} response = client.post( path='/images/replaceBackground', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, files={ 'image': open('image.png', 'rb'), }, body={ 'prompt': 'winter', }, ) print(response['data'][0]['url']) ``` ### Parameters Body of a request should contain an image file in PNG, JPG or WEBP format and parameters passed as content type `'multipart/form-data'`. The image must be no more than 5 MB in size, have resolution no more than 16 MP, max dimension no more than 4096 pixels and min dimension no less than 256 pixels. | Parameter | Type | Description | Compatibility | | ----------------- | -------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------ | | image (required) | file | An image to modify, must be less than 5 MB in size, have resolution less than 16 MP, and max dimension less than 4096 pixels. | | | prompt (required) | string | A text description of areas to change. | refer to [Appendix](/api-reference/appendix#maximum-prompt-length) | | n | integer or null, default is 1 | The number of images to generate, must be between 1 and 6. | | | model | string or null, default is `recraftv3` | The model to use for image generation. | Must be one of: `recraftv3`, `recraftv3_vector` | | style | string or null | The style of the generated images, refer to [Styles](/api-reference/styles). | V3 styles only | | style\_id | UUID or null | Use a style as visual reference, refer to [Styles](/api-reference/styles). | V3 styles only | | response\_format | string or null, default is `url` | The format in which the generated images are returned. Must be one of `url` or `b64_json`. | | | negative\_prompt | string or null | A text description of undesired elements on an image. | | | text\_layout | Array of objects or null | Refer to [Text Layout](#text-layout). | | | controls | object or null | A set of custom parameters to tweak generation process, refer to [Controls](#controls). | | ## Generate background Generate Background operation generates a background for a given image, based on a prompt and a mask that specifies the regions to fill. ***Note***: This operation is available with Recraft V3, Recraft V3 Vector models only. ```javascript theme={null} POST https://external.api.recraft.ai/v1/images/generateBackground ``` ### Example ```python theme={null} response = client.post( path='/images/generateBackground', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, files={ 'image': open('image.png', 'rb'), 'mask': open('mask.png', 'rb'), }, body={ 'prompt': 'winter', }, ) print(response['data'][0]['url']) ``` ### Parameters Body of a request should contain an image file and a mask file, both in PNG, JPG or WEBP format, and parameters passed as content type `'multipart/form-data'`. The image must be no more than 5 MB in size, have resolution no more than 16 MP, max dimension no more than 4096 pixels and min dimension no less than 256 pixels. | Parameter | Type | Description | Compatibility | | ----------------- | -------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------ | | image (required) | file | An image to modify, must be less than 5 MB in size, have resolution less than 16 MP, and max dimension less than 4096 pixels. | | | mask (required) | file | An image encoded in **grayscale color mode**, used to define the specific regions of an image that need modification. The white pixels represent the parts of the image that will be inpainted, while black pixels indicate the parts of the image that will remain unchanged. Should have exactly the same size as the image. Each pixel of the image should be either pure black (value `0`) or pure white (value `255`). | | | prompt (required) | string | A text description of areas to change. | refer to [Appendix](/api-reference/appendix#maximum-prompt-length) | | n | integer or null, default is 1 | The number of images to generate, must be between 1 and 6. | | | model | string or null, default is `recraftv3` | The model to use for image generation. | Must be one of: `recraftv3`, `recraftv3_vector` | | style | string or null | The style of the generated images, refer to [Styles](/api-reference/styles). | V3 styles only | | style\_id | UUID or null | Use a style as visual reference, refer to [Styles](/api-reference/styles). | V3 styles only | | response\_format | string or null, default is `url` | The format in which the generated images are returned. Must be one of `url` or `b64_json`. | | | negative\_prompt | string or null | A text description of undesired elements on an image. | | | text\_layout | Array of objects or null | Refer to [Text Layout](#text-layout). | | | controls | object or null | A set of custom parameters to tweak generation process, refer to [Controls](#controls). | | ## Vectorize image Converts a given raster image to SVG format. ```javascript theme={null} POST https://external.api.recraft.ai/v1/images/vectorize ``` ### Example ```javascript theme={null} response = client.post( path='/images/vectorize', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, files={'file': open('image.png', 'rb')}, ) print(response['image']['url']) ``` ```text theme={null} "svg_compression": "on/off" (default off) "limit_num_shapes": "on/off" (default off) "max_num_shapes": int (not set by default) ``` ### **Parameters** Body of a request should be a file in PNG, JPG or WEBP format and parameters passed as content type `'multipart/form-data'`. The image must be no more than 5 MB in size, have resolution no more than 16 MP, max dimension no more than 4096 pixels and min dimension no less than 256 pixels. | Parameter | Type | Description | | ---------------- | -------------------------------- | ------------------------------------------------------------------------------------------ | | response\_format | string or null, default is `url` | The format in which the generated images are returned. Must be one of `url` or `b64_json`. | ## Remove background Removes background of a given raster image. ```javascript theme={null} POST https://external.api.recraft.ai/v1/images/removeBackground ``` ### Example ```javascript theme={null} response = client.post( path='/images/removeBackground', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, files={'file': open('image.png', 'rb')}, ) print(response['image']['url']) ``` ### **Parameters** Body of a request should be a file in PNG, JPG or WEBP format and parameters passed as content type `'multipart/form-data'`. The image must be no more than 5 MB in size, have resolution no more than 16 MP, max dimension no more than 4096 pixels and min dimension no less than 256 pixels. | Parameter | Type | Description | | ---------------- | -------------------------------- | ------------------------------------------------------------------------------------------ | | response\_format | string or null, default is `url` | The format in which the generated images are returned. Must be one of `url` or `b64_json`. | ## Crisp upscale Enhances a given raster image using ‘crisp upscale’ tool, increasing image resolution, making the image sharper and cleaner. ```javascript theme={null} POST https://external.api.recraft.ai/v1/images/crispUpscale ``` ### Example ```javascript theme={null} response = client.post( path='/images/crispUpscale', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, files={'file': open('image.png', 'rb')}, ) print(response['image']['url']) ``` ### Request body Body of a request should be a file in PNG, JPG or WEBP format and parameters passed as content type `multipart/form-data`. The image must be no more than 5 MB in size, have resolution no more than 4 MP, max dimension no more than 4096 pixels and min dimension no less than 32 pixels. | Parameter | Type | Description | | ---------------- | -------------------------------- | ------------------------------------------------------------------------------------------ | | response\_format | string or null, default is `url` | The format in which the generated images are returned. Must be one of `url` or `b64_json`. | ## Creative upscale Enhances a given raster image using ‘creative upscale’ tool, boosting resolution with a focus on refining small details and faces. ```javascript theme={null} POST https://external.api.recraft.ai/v1/images/creativeUpscale ``` ### Example ```javascript theme={null} response = client.post( path='/images/creativeUpscale', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, files={'file': open('image.png', 'rb')}, ) print(response['image']['url']) ``` ### Request body Body of a request should be a file in PNG, JPG or WEBP format and parameters passed as content type `multipart/form-data`. The image must be no more than 5 MB in size, have resolution no more than 16 MP, max dimension no more than 4096 pixels and min dimension no less than 256 pixels. | Parameter | Type | Description | | ---------------- | -------------------------------- | ------------------------------------------------------------------------------------------ | | response\_format | string or null, default is `url` | The format in which the generated images are returned. Must be one of `url` or `b64_json`. | ## Erase region Erases a region of a given raster image following a given mask, where white pixels represent the regions to erase, and black pixels indicate the areas to keep intact. ```javascript theme={null} POST https://external.api.recraft.ai/v1/images/eraseRegion ``` ### Example ```javascript theme={null} response = client.post( path='/images/eraseRegion', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, files={'image': open('image.png', 'rb'), 'mask': open('mask.png', 'rb')}, ) print(response['image']['url']) ``` ### Request body Body of a request should contain a file and a mask, both in PNG, JPG or WEBP format, and parameters passed as content type `multipart/form-data`. The images must be no more than 5 MB in size, have resolution no more than 4 MP, max dimension no more than 4096 pixels and min dimension no less than 32 pixels. The mask and image must have the same dimensions. | Parameter | Type | Description | | ---------------- | -------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | image (required) | file | An image to modify, must be less than 5 MB in size, have resolution less than 16 MP, and max dimension less than 4096 pixels. | | mask (required) | file | An image encoded in **grayscale color mode**, used to define the specific regions of the image to be erased. The white pixels represent the parts of the image that will be erased, while black pixels indicate the parts of the image that will remain unchanged. Should have exactly the same size as the image. Each pixel of the image should be either pure black (value `0`) or pure white (value `255`). | | response\_format | string or null, default is `url` | The format in which the generated images are returned. Must be one of `url` or `b64_json`. | ## Remix image Remix generates new versions of an image while keeping the overall concept intact. Each remix introduces slight differences in elements like character appearance, object position, or scene composition. The Remix function does not use a prompt. Instead, it uses the visual content of the original image to generate alternatives in different aspect ratios. ```javascript theme={null} POST https://external.api.recraft.ai/v1/images/variateImage ``` ### Example ```python theme={null} response = client.post( path='/images/variateImage', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, files={'image': open('image.png', 'rb')}, body={'size': '1024x1024'} ) print(response['data'][0]['url']) "svg_compression": "on/off" (default off) "limit_num_shapes": "on/off" (default off) "max_num_shapes": int (not set by default) ``` ### Parameters The request body must be a file in PNG, JPG, or WEBP format, submitted with the content type 'multipart/form-data'. The image must not exceed 5 MB in size, with a maximum resolution of 16 MP, a maximum dimension of 4096 px, and a minimum dimension of 256 px. | Parameter | Type | Description | | ---------------- | -------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------- | | image (required) | file | The input image in PNG, WEBP or JPEG format. | | size (required) | string | The size of the generated images in `WxH` or `w:h` format, supported values are published in [Appendix](/api-reference/appendix#list-of-image-sizes). | | n | integer or null, default is 1 | Number of variations to generate \[1–6]. | | random\_seed | string or null | Optional random seed for reproducibility. | | response\_format | string or null, default is `url` | The format in which the generated images are returned. Must be one of `url` or `b64_json`. | | image\_format | string or null | Format of the output image. Must be one of: `png`, `webp`. | ## Get user information Returns information of the current user including credits balance. ```javascript theme={null} GET https://external.api.recraft.ai/v1/users/me ``` ### Example ```python theme={null} response = client.get(path='/users/me', cast_to=object) print(response) ``` ### Output ```javascript theme={null} { "credits": 1000, "email": "test@example.com", "id": "c18a1988-45e7-4c00-82c4-4ad7d3dbce3a", "name": "Recraft Test" } ``` ## Auxiliary ### Controls The generation process can be adjusted with a number of tweaks. | Parameter | Type | Description | Compatibility | | ----------------- | -------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------------- | | colors | array of color definitions | An array of preferable colors. | All models | | background\_color | color definition | Use the given color as a desired background color. | All models | | artistic\_level | integer or null | Defines the artistic tone of your image. At a simple level, the person looks straight at the camera in a static and clean style. Dynamic and eccentric levels introduce movement and creativity. The value should be in the range `[0..5]`. | V3 models only | | no\_text | bool | Do not embed text layouts. | V3 models only | #### Colors Color type is defined as an object with the following fields | Parameter | Description | | -------------- | ---------------------------------------------------------------------------- | | rgb (required) | An array of 3 integer values in range of `0...255` defining RGB color model. | **Example** ```python theme={null} response = client.images.generate( prompt='race car on a track', extra_body={ 'controls': { 'colors': [ {'rgb': [0, 255, 0]} ] } } ) print(response.data[0].url) ``` ### Text Layout Text layout is used to define spatial and textual attributes for individual text elements. Each text element consists of an individual word and its bounding box (bbox). This feature is supported by Recraft V3 and Recraft V3 Vector models only. | Parameter | Description | | --------------- | ------------------------------------------------------------------------------------------------------------- | | text (required) | A single word containing only supported characters. | | bbox (required) | A bounding box representing a 4-angled polygon. Each point in the polygon is defined by relative coordinates. | **Bounding box**: The bounding box (bbox) is a list of 4 points representing a 4-angled figure (not necessarily a rectangle). Each point specifies its coordinates relative to the layout dimensions, where (0, 0) is the top-left corner, (1, 1) is the bottom-right corner. **Coordinates**: Coordinates are **relative** to the layout dimensions. Coordinates can extend beyond the \[0, 1] range, such values indicate that the shape will be cropped. **Points**: The bounding box must always have **exactly 4 points**. Each point is an array of two floats \[x, y] representing the relative position. **Supported characters** The `text` field must contain a single word composed only of the following characters: ```text theme={null} ! " # $ % & ' ( ) * + , - . / 0 1 2 3 4 5 6 7 8 9 : ; < > ? @ A B C D E F G H I J K L M N O P Q R S T U V W X Y Z _ { } Ø Đ Ħ Ł Ŋ Ŧ Α Β Ε Ζ Η Ι Κ Μ Ν Ο Ρ Τ Υ Χ І А В Е К М Н О Р С Т У Х ß ẞ ``` Any character not listed above will result in validation errors. **Example** ```python theme={null} response = client.images.generate( prompt="cute red panda with a sign", style="Illustration", extra_body={ "text_layout": [ { "text": "Recraft", "bbox": [[0.3, 0.45], [0.6, 0.45], [0.6, 0.55], [0.3, 0.55]], }, { "text": "AI", "bbox": [[0.62, 0.45], [0.70, 0.45], [0.70, 0.55], [0.62, 0.55]], }, ] }, ) print(response.data[0].url) ``` # Examples Source: https://www.recraft.ai/docs/api-reference/examples Generate AI images using cURL or Python and create your own styles programmatically. ### Generate a raster image using Recraft V4 model ```bash generate_recraftv4.sh theme={null} curl https://external.api.recraft.ai/v1/images/generations \ -H "Content-Type: application/json" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -d '{ "prompt": "two race cars on a track", "model": "recraftv4" }' ``` ```python generate_recraftv4.py theme={null} from openai import OpenAI client = OpenAI(base_url='https://external.api.recraft.ai/v1', api_key=_RECRAFT_API_TOKEN) response = client.images.generate( prompt='two race cars on a track', model='recraftv4' ) print(response.data[0].url) ``` ### Generate a vector image using Recraft V4 Vector model ```bash generate_recraftv4_vector.sh theme={null} curl https://external.api.recraft.ai/v1/images/generations \ -H "Content-Type: application/json" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -d '{ "prompt": "cat on a mat", "model": "recraftv4_vector" }' ``` ```python generate_recraftv4_vector.py theme={null} from openai import OpenAI client = OpenAI(base_url='https://external.api.recraft.ai/v1', api_key=_RECRAFT_API_TOKEN) response = client.images.generate( prompt='cat on a mat', model='recraftv4_vector' ) print(response.data[0].url) ``` ### Generate a realistic image by Recraft V3 with specific size ```bash generate_with_size.sh theme={null} curl https://external.api.recraft.ai/v1/images/generations \ -H "Content-Type: application/json" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -d '{ "prompt": "red point siamese cat", "model": "recraftv3", "style": "Photorealism", "size": "1280x1024" }' ``` ```python generate_with_size.py theme={null} from openai import OpenAI client = OpenAI(base_url='https://external.api.recraft.ai/v1', api_key=_RECRAFT_API_TOKEN) response = client.images.generate( prompt='red point siamese cat', model='recraftv3', style='Photorealism', size='1280x1024', ) print(response.data[0].url) ``` ### Generate a digital illustration by Recraft V3 with specific style ```bash generate_digital_illustration.sh theme={null} curl https://external.api.recraft.ai/v1/images/generations \ -H "Content-Type: application/json" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -d '{ "prompt": "a monster with lots of hands", "style": "Hand-drawn" }' ``` ```python generate_digital_illustration.py theme={null} from openai import OpenAI client = OpenAI(base_url='https://external.api.recraft.ai/v1', api_key=_RECRAFT_API_TOKEN) response = client.images.generate( prompt='a monster with lots of hands', style='Hand-drawn', ) print(response.data[0].url) ``` ### Image to image by Recraft V3 with digital illustration style ```bash image_to_image.sh theme={null} curl -X POST https://external.api.recraft.ai/v1/images/imageToImage \ -H "Content-Type: multipart/form-data" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -F "image=@image.png" \ -F "prompt=winter" \ -F "strength=0.2" \ -F "style=Illustration" ``` ```python image_to_image.py theme={null} from openai import OpenAI client = OpenAI(base_url='https://external.api.recraft.ai/v1', api_key=_RECRAFT_API_TOKEN) response = client.post( path='/images/imageToImage', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, files={ 'image': open('image.png', 'rb'), }, body={ 'prompt': 'winter', 'strength': 0.2, 'style': 'Illustration', }, ) print(response['data'][0]['url']) ``` ### Inpaint an image by Recraft V3 with style Enterprise ```bash inpaint.sh theme={null} curl -X POST https://external.api.recraft.ai/v1/images/inpaint \ -H "Content-Type: multipart/form-data" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -F "prompt=moon" \ -F "style=Enterprise" \ -F "image=@image.png" -F "mask=@mask.png" ``` ```python inpaint.py theme={null} from openai import OpenAI client = OpenAI(base_url='https://external.api.recraft.ai/v1', api_key=_RECRAFT_API_TOKEN) response = client.post( path='/images/inpaint', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, files={ 'image': open('image.png', 'rb'), 'mask': open('mask.png', 'rb'), }, body={ 'style': 'Enterprise', 'prompt': 'moon', }, ) print(response['data'][0]['url']) ``` ### Create own Recraft V3 style by uploading reference images and use them for generation ```bash create_style_and_generate.sh theme={null} curl -X POST https://external.api.recraft.ai/v1/styles \ -H "Content-Type: multipart/form-data" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -F "style=digital_illustration" \ -F "file=@image.png" # response: {"id":"095b9f9d-f06f-4b4e-9bb2-d4f823203427"} curl https://external.api.recraft.ai/v1/images/generations \ -H "Content-Type: application/json" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -d '{ "prompt": "wood potato masher", "style_id": "095b9f9d-f06f-4b4e-9bb2-d4f823203427" }' ``` ```python create_style_and_generate.py theme={null} from openai import OpenAI client = OpenAI(base_url='https://external.api.recraft.ai/v1', api_key=_RECRAFT_API_TOKEN) style = client.post( path='/styles', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, body={'style': 'digital_illustration'}, files={'file': open('image.png', 'rb')}, ) print(style['id']) response = client.images.generate( prompt='wood potato masher', extra_body={'style_id': style['id']}, ) print(response.data[0].url) ``` ### Vectorize an image in PNG format ```bash vectorize.sh theme={null} curl -X POST https://external.api.recraft.ai/v1/images/vectorize \ -H "Content-Type: multipart/form-data" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -F "file=@image.png" ``` ```python vectorize.py theme={null} from openai import OpenAI client = OpenAI(base_url='https://external.api.recraft.ai/v1', api_key=_RECRAFT_API_TOKEN) response = client.post( path='/images/vectorize', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, files={'file': open('image.png', 'rb')}, ) print(response['image']['url']) ``` ### Remove background from a PNG image, get the result in B64 JSON ```bash remove_background_b64.sh theme={null} curl -X POST https://external.api.recraft.ai/v1/images/removeBackground \ -H "Content-Type: multipart/form-data" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -F "response_format=b64_json" \ -F "file=@image.png" ``` ```python remove_background_b64.py theme={null} from openai import OpenAI client = OpenAI(base_url='https://external.api.recraft.ai/v1', api_key=_RECRAFT_API_TOKEN) response = client.post( path='/images/removeBackground', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, body={'response_format': 'b64_json'}, files={'file': open('image.png', 'rb')}, ) print(response['image']['url']) ``` ### Run crisp upscale tool for a PNG image, get the result in B64 JSON ```bash crisp_upscale_b64.sh theme={null} curl -X POST https://external.api.recraft.ai/v1/images/crispUpscale \ -H "Content-Type: multipart/form-data" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -F "response_format=b64_json" \ -F "file=@image.png" ``` ```python crisp_upscale_b64.py theme={null} from openai import OpenAI client = OpenAI(base_url='https://external.api.recraft.ai/v1', api_key=_RECRAFT_API_TOKEN) response = client.post( path='/images/crispUpscale', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, body={'response_format': 'b64_json'}, files={'file': open('image.png', 'rb')}, ) print(response['image']['url']) ``` ### Run creative upscale tool for a PNG image ```bash creative_upscale.sh theme={null} curl -X POST https://external.api.recraft.ai/v1/images/creativeUpscale \ -H "Content-Type: multipart/form-data" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -F "file=@image.png" ``` ```python creative_upscale.py theme={null} from openai import OpenAI client = OpenAI(base_url='https://external.api.recraft.ai/v1', api_key=_RECRAFT_API_TOKEN) response = client.post( path='/images/creativeUpscale', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, files={'file': open('image.png', 'rb')}, ) print(response['image']['url']) ``` ### Variate PNG image, get the result in WEBP format ```bash variate_image.sh theme={null} curl -X POST https://external.api.recraft.ai/v1/images/variateImage \ -H "Content-Type: multipart/form-data" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -F "response_format=url" \ -F "size=1024x1024" \ -F "n=1" \ -F "seed=13191922" \ -F "image_format=webp" \ -F "file=@image.png" ``` ```python variate_image.py theme={null} from openai import OpenAI client = OpenAI(base_url='https://external.api.recraft.ai/v1', api_key=_RECRAFT_API_TOKEN) response = client.post( path='/images/variateImage', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, body={"size": "1024x1024", "n": 1, "response_format": "url", "seed": 13191922, "image_format": "webp"}, files={'file': open('image.png', 'rb')}, ) print(response['data'][0]["url"]) ``` # Getting started Source: https://www.recraft.ai/docs/api-reference/getting-started Welcome to Recraft image generation and editing API. Learn the basics of the Recraft API, including raster and vector image generation, style creation, image generation in your brand style and colors, image vectorization, and background removal. **Authenticate and interact with our API in a matter of minutes.** ### Models Recraft has developed three generations of proprietary models: * Recraft V4, Recraft V4 Vector, Recraft V4 Pro, Recraft V4 Pro Vector * Recraft V3 (aka Red Panda), Recraft V3 Vector * Recraft V2 (aka Recraft 20B), Recraft V2 Vector Recraft V4 models, released in February 2026, represent a ground-up rebuild and our most advanced generation to date. Focused on what matters for real work — design taste, prompt accuracy, and output quality at any size — V4 models deliver true design vision in composition, lighting, textures, and overall feel, along with the world's only production-grade vector generation. Available in two versions: V4 (1MP) for everyday work and fast iteration, and V4 Pro (4MP) for print-ready assets and large-scale use. Both share the same creative capabilities, design taste, and art-directed professional results. Recraft V3 models, released in October 2024, were new state-of-the-art models trained from scratch that introduced major advances in photorealism and text rendering. V3 ranked first on the Hugging Face Text-to-Image leaderboard for five consecutive months. Recraft V2 models, released in February 2024, were the first AI models built specifically for designers. V2 enabled the creation of both raster and vector images with anatomical accuracy, consistent brand styling, and precise iteration control. All eight models from V2, V3, and V4 generations are available in the Recraft API and within the Recraft web and mobile applications. ## Authentication We use [Bearer](https://swagger.io/docs/specification/authentication/bearer-authentication/) API tokens for authentication. To access your API key, log in to Recraft, [enter your profile](https://app.recraft.ai/profile/api) and hit 'Generate' (available only if your API units balance is above zero). All requests should include your API key in an Authorization HTTP header as follows:\ \ `Authorization: Bearer RECRAFT_API_TOKEN` A user may create multiple API tokens; however, all tokens share the same API units balance. ## Swagger Interactive API documentation is available via [Swagger](https://external.api.recraft.ai/doc/#/). You can explore all available endpoints, test API calls, and view request/response schemas directly in your browser. ### REST / Python Library The Recraft API adheres to REST principles, allowing you to interact using any utilities (e.g., curl), programming languages, or libraries of your choice. One of the easiest of available alternatives is [OpenAI Python library](https://github.com/openai/openai-python) which is also compatible with Recraft API, but it’s important to remember that not all parameters/options are supported or implemented. Additionally, some parameters may have different meanings, or they may be quietly ignored if they are not applicable to the Recraft API. Future examples will be shown using that library, for example, once installed, you can use the following code to be authenticated: ```python theme={null} from openai import OpenAI client = OpenAI( base_url='https://external.api.recraft.ai/v1', api_key=, ) ``` # Pricing Source: https://www.recraft.ai/docs/api-reference/pricing ## API unit package pricing The following are the API Unit packages available from Recraft for the use of the Recraft API Service. API Unit packages must be purchased in advance and all API Unit packages are non-cancellable and non-refundable. Any number of unit packages can be bought, and purchased unit packages do not expire. | Price | API units | | ---------- | --------- | | USD \$1.00 | 1,000 | ## API unit charges for API services The following are the service charges (in API Units) for use of the API Services. API Units will be automatically deducted from Member’s pre-purchased API Unit package. | Service description | Cost (USD) | Cost (API units) | Billing Basis | | ----------------------------------------------- | ---------- | ---------------- | ------------- | | Raster image generation – Recraft V4 Pro | \$0.25 | 250 | Per image | | Raster image generation – Recraft V4 | \$0.04 | 40 | Per image | | Raster image generation – Recraft V3 | \$0.04 | 40 | Per image | | Raster image generation – Recraft V2 | \$0.022 | 22 | Per image | | Vector image generation – Recraft V4 Pro Vector | \$0.30 | 300 | Per image | | Vector image generation – Recraft V4 Vector | \$0.08 | 80 | Per image | | Vector image generation – Recraft V3 Vector | \$0.08 | 80 | Per image | | Vector image generation – Recraft V2 Vector | \$0.044 | 44 | Per image | | Raster image to image – Recraft V3 | \$0.04 | 40 | Per image | | Vector image to image – Recraft V3 Vector | \$0.08 | 80 | Per image | | Raster image inpainting – Recraft V3 | \$0.04 | 40 | Per image | | Vector image inpainting – Recraft V3 Vector | \$0.08 | 80 | Per image | | Replace raster background – Recraft V3 | \$0.04 | 40 | Per image | | Replace vector background – Recraft V3 Vector | \$0.08 | 80 | Per image | | Generate raster background – Recraft V3 | \$0.04 | 40 | Per image | | Generate vector background – Recraft V3 Vector | \$0.08 | 80 | Per image | | Image style creation | \$0.04 | 40 | Per request | | Image vectorization | \$0.01 | 10 | Per request | | Image background removal | \$0.01 | 10 | Per request | | Crisp upscale | \$0.004 | 4 | Per request | | Creative upscale | \$0.25 | 250 | Per request | | Erase region | \$0.002 | 2 | Per request | | Variate image | \$0.04 | 40 | Per request | # Styles Source: https://www.recraft.ai/docs/api-reference/styles **Note:** Styles are **not yet supported for V4 models**. Styles define the visual appearance and aesthetic of generated images, including textures, visual effects, colors, composition, and overall artistic feel. Recraft offers both predefined curated styles and the ability to create custom styles. Recraft styles are organized into groups: realistic, digital illustrations, vector illustrations, icons, etc. Please refer to the examples below for visual references. To generate an image, you may provide a specific style name, such as `Recraft V3 Raw`, `Photorealism`, or `Vector art`. If a style name is supported by multiple models, the API defaults to Recraft V3 (if supported, to Recraft V2 otherwise). To use a specific version, explicitly pair the style with the model parameter (e.g., `Photorealism` with `recraftv3` or `Vector art` with `recraftv2_vector`). Refer to the [List of Styles](/api-reference/styles#list-of-styles) for compatibility across V2 / V3 models. Examples: Images of style `Photorealism` are expected to look like just ordinary photographs made with a digital camera or a smartphone or a film camera. Photorealism Images of style `Illustration` are pictures drawn by hand or using computers - virtually everything except photos and vector illustrations. The most crucial difference from `Photorealism` is that illustrations possess simplified textures (like in 3D-rendered or manually drawn images) - or they are stylized in a certain creative way. The difference from `Vector art` is that `Illustration` allow for more complex color transitions, shades, fine textures. ![](https://cdn.prod.website-files.com/655727fe69827d9a402de12c/679b8318c6f6ecaf608531da_unicorn%20\(8\).png) ![](https://cdn.prod.website-files.com/655727fe69827d9a402de12c/679b82f54ef9617939df9d8e_motorcycle--going-fast.png) Images of style `Vector art` are expected to look like those drawn using vector graphics (see [Wikipedia](https://en.wikipedia.org/wiki/Vector_graphics)). Usually, they use only a few different colors at once, shapes are filled with flat colors or simple color gradients. Shapes of objects can be arbitrarily complex. ![](https://cdn.prod.website-files.com/655727fe69827d9a402de12c/679b82fb365f45ed1a743d43_a-man-working-in-his-workshop--standing-straight-a%20\(1\)photorealism-exp1-1.png) ![](https://cdn.prod.website-files.com/655727fe69827d9a402de12c/679b830165533469004467d9_create-a-minimal-character-illustration-of-a-resea.png) Images of style `Icon` are small digital images or symbols used in the graphical user interface. They are designed to be simple and recognizable at small sizes, often visually summarizing the action or object they stand for, or they can act as the visual identity for an app or a website and are crucial in branding. ![](https://cdn.prod.website-files.com/655727fe69827d9a402de12c/679b830d1720fd793595cf40_Group%204291.png) ![](https://cdn.prod.website-files.com/655727fe69827d9a402de12c/679b831e6ee14a2f933f112c_Group%204254.png) `Photorealism` and `Illustration` images are generated in raster formats such as PNG, WEBP, or JPG. These formats use pixels to represent details, making them ideal for photos and complex artwork. On the other hand, `Vector art` and `Icon` are generated in the SVG format. Unlike raster images, SVGs are made of mathematical paths and shapes, allowing them to scale to any size without losing quality. ### Using of styles The `style` parameter accepts the name of any curated style provided by Recraft. Examples include but not limited to: * Recraft V4, Recraft V4 Vector, Recraft V4 Pro, Recraft V4 Vector Pro: **styles are not supported** * Recraft V3: [`Recraft V3 Raw`](https://www.recraft.ai/styles/b9af225b-6b98-43d0-a4b3-16343641e852), [`Photorealism`](https://www.recraft.ai/styles/6b98805f-9342-4d1e-9b53-bcb6b6d20ea1), [`Illustration`](https://www.recraft.ai/styles/3eb71874-5e13-4386-a74b-e42119186b0a), [`Enterprise`](https://www.recraft.ai/styles/e4eb5a3f-1b8b-4f5c-bada-bf6c2451cb17), [`Hand-drawn`](https://www.recraft.ai/styles/17027120-cf06-465f-a4d0-2e44ad8dee0d), [`Punk Graphic`](https://www.recraft.ai/styles/84789d26-1153-4faf-a1d1-c3e94d7a0035), etc * Recraft V3 Vector: [`Vector art`](https://www.recraft.ai/styles/773902ac-683a-4042-86a3-fe714eee345a), [`Line art`](https://www.recraft.ai/styles/19942671-ffa4-4867-9bbe-f5e4503dcc93), [`Engraving`](https://www.recraft.ai/styles/ce1f9161-a8dc-488f-9ff1-3f41082887b5), etc * Recraft V2: [`Photorealism`](https://www.recraft.ai/styles/5a99ae89-3f14-433b-abb2-a1d8074b1ab1), [`Illustration`](https://www.recraft.ai/styles/b8772060-8802-4f3c-b051-73349b1017a5), [`3D render`](https://www.recraft.ai/styles/4659b2d6-a580-4287-849e-96190d7dbf30), [`Kawaii`](https://www.recraft.ai/styles/f8f1bc1c-1ae7-43ea-a36f-c81049feee01), etc * Recraft V2 Vector: [`Vector art`](https://www.recraft.ai/styles/807ce9bc-7e36-4efa-a6cb-11cda1cecd19), [`Icon`](https://www.recraft.ai/styles/645cb7ca-4be3-4d90-9f16-e9c26f6812f1), [`Doodle`](https://www.recraft.ai/styles/a2ba0c04-e40d-4210-8ed2-46d3537c546c), etc For the complete list of available styles organized by model and category, see the [List of Styles](/api-reference/styles#list-of-styles). Some styles with identical names exist across multiple models (e.g., `Photorealism` is available in Recraft V2, and Recraft V3 models). Use the `model` parameter to explicitly specify which model version to use. The `style_id` parameter accepts: * style IDs created via the API; * style IDs from the Recraft web platform, accessible if : * you own the style; * the style is publicly available; * the style was explicitly shared to your account. To obtain a style ID from the web platform, open any style in the Styles panel and click the three-dot menu to copy the style ID. The `style` and `style_id` parameters cannot be used together. Specify one or the other, not both. If neither parameter is provided, the default style for the selected model will be used: * Recraft V3: [`Recraft V3 Raw`](https://www.recraft.ai/styles/b9af225b-6b98-43d0-a4b3-16343641e852); * Recraft V3 Vector: [`Vector art`](https://www.recraft.ai/styles/773902ac-683a-4042-86a3-fe714eee345a); * Recraft V2: [`Photorealism`](https://www.recraft.ai/styles/5a99ae89-3f14-433b-abb2-a1d8074b1ab1); * Recraft V2 Vector: [`Vector art`](https://www.recraft.ai/styles/807ce9bc-7e36-4efa-a6cb-11cda1cecd19). Model compatibility: * custom styles created on the web platform are compatible with the model specified during creation; * custom styles created via the API are compatible with Recraft V3 and Recraft V3 Vector models only. ### List of styles To generate an image, you may provide a specific style name, such as `Recraft V3 Raw`, `Photorealism`, or `Vector art`. If a style name is supported by multiple models, the API defaults to Recraft V3 (if supported, to Recraft V2 otherwise). To use a specific version, explicitly pair the style with the model parameter (e.g., `Photorealism` with `recraftv3` or `Vector art` with `recraftv2_vector`). #### Recraft V3 * Photorealistic styles: [`Photorealism`](https://www.recraft.ai/styles/6b98805f-9342-4d1e-9b53-bcb6b6d20ea1), [`Enterprise`](https://www.recraft.ai/styles/e4eb5a3f-1b8b-4f5c-bada-bf6c2451cb17), [`Natural light`](https://www.recraft.ai/styles/90cf050a-5985-4ca8-b255-d370831701ca), [`Studio photo`](https://www.recraft.ai/styles/b8f16786-5794-47fd-b2b3-5f607d919d1c), [`HDR`](https://www.recraft.ai/styles/d0c61ed6-5f6c-4332-af27-e6a5fc2bde64), [`Hard flash`](https://www.recraft.ai/styles/41501ef8-26a6-4d1c-946c-d8b659ff11c3), [`Motion blur`](https://www.recraft.ai/styles/ca56e01b-a128-477f-b2f3-13a7191c8e91), [`Black & white`](https://www.recraft.ai/styles/7babd38f-dc87-4ccf-99de-d8834f8900f5), [`Evening light`](https://www.recraft.ai/styles/d92bef4c-605e-451d-a93f-a8d78f14effa), [`Faded Nostalgia`](https://www.recraft.ai/styles/889bee89-edce-4104-b572-df38c58e8659), [`Forest life`](https://www.recraft.ai/styles/51536681-7ee0-4604-bbb3-3a9e45e2b59f), [`Mystic Naturalism`](https://www.recraft.ai/styles/3ab91eeb-203f-4e00-865b-1e64c7854f08), [`Natural Tones`](https://www.recraft.ai/styles/45f34903-6da9-4468-966d-11314d845b41), [`Organic Calm`](https://www.recraft.ai/styles/06401905-a828-4652-a962-7026693c8cf7), [`Real-Life Glow`](https://www.recraft.ai/styles/336eabc6-da87-4a1e-a567-c3dd17bf5e79), [`Retro Realism`](https://www.recraft.ai/styles/2a4176f2-be17-49c5-b8b7-7a3bdd11a277), [`Retro Snapshot`](https://www.recraft.ai/styles/226190a0-5c43-4e0c-98a0-1a4ba63800cd), [`Urban Drama`](https://www.recraft.ai/styles/c61a727c-a9a8-40a5-a289-fdaf3a59eae1), [`Village Realism`](https://www.recraft.ai/styles/ba5e9c0c-4a20-492d-924d-a758b52b102d), [`Warm Folk`](https://www.recraft.ai/styles/db207d41-37bf-46ae-9e12-3201167c4cbe), [`Product photo`](https://www.recraft.ai/styles/57ed072c-3e0e-4df5-baba-83a6e87048b8); * Illustration styles: [`Illustration`](https://www.recraft.ai/styles/3eb71874-5e13-4386-a74b-e42119186b0a), [`Hand-drawn`](https://www.recraft.ai/styles/17027120-cf06-465f-a4d0-2e44ad8dee0d), [`Grain`](https://www.recraft.ai/styles/245b724f-8a24-4bde-aba4-21d41295074d), [`Bold Sketch`](https://www.recraft.ai/styles/6319466c-2bda-4b13-88ba-3e4e82aba55e), [`Pencil sketch`](https://www.recraft.ai/styles/8a5273c5-12cf-48de-8217-0f879b881398), [`Retro Pop`](https://www.recraft.ai/styles/a56d7265-ed27-4e0d-827b-2692d0b46fba), [`Clay`](https://www.recraft.ai/styles/f5145ada-d0d5-46d9-adda-d38fdf0b80d9), [`Risograph`](https://www.recraft.ai/styles/7be82d9c-ad40-483e-b235-389043675ad2), [`Color engraving`](https://www.recraft.ai/styles/75ee357c-ff80-4fbe-99ae-44d3310155fe), [`Pixel art`](https://www.recraft.ai/styles/aa32a571-eeeb-48bc-bc51-5792e9f49eb4), [`Antiquarian`](https://www.recraft.ai/styles/2b6d8781-7ab6-43dc-b870-891cc072246d), [`Bold fantasy`](https://www.recraft.ai/styles/1703ae77-b491-47e2-a25d-9dd6c28c2ffa), [`Child book`](https://www.recraft.ai/styles/ecd1de2b-02b7-43ce-9780-2ea91deb38f4), [`Cover`](https://www.recraft.ai/styles/6e403112-b52d-408b-8f78-539657229ade), [`Crosshatch`](https://www.recraft.ai/styles/7cccbbd2-c282-4ef6-a37f-889d1c503b74), [`Digital engraving`](https://www.recraft.ai/styles/f0999093-d0dc-4718-adba-a038a60ff6b2), [`Expressionism`](https://www.recraft.ai/styles/e2d8e964-ea62-43f4-a269-a47dd8728fe5), [`Freehand details`](https://www.recraft.ai/styles/eac001dc-5c63-4e6b-8d1f-390cbe5d2040), [`Grain 2.0`](https://www.recraft.ai/styles/824645a6-dc5b-4ad2-95c2-90bb6db5fbe1), [`Graphic intensity`](https://www.recraft.ai/styles/88b9e740-728c-4d00-9725-6d820859f6b7), [`Hard Comics`](https://www.recraft.ai/styles/98cef725-f770-498b-a90e-2eb4ab03cddb), [`Long shadow`](https://www.recraft.ai/styles/9a3edce0-424f-49c5-b372-61b0f589426c), [`Modern Folk`](https://www.recraft.ai/styles/6aa86245-e03f-432d-8b57-5fe5bfdb60cf), [`Multicolor`](https://www.recraft.ai/styles/da5cc27d-0389-4b5c-8a0d-fe5b76937214), [`Neon Calm`](https://www.recraft.ai/styles/1feb869d-edcc-4b6b-b6d0-413e9560f2dd), [`Noir`](https://www.recraft.ai/styles/956a5ac8-401f-4e5f-a034-ec458c371598), [`Nostalgic pastel`](https://www.recraft.ai/styles/4ce80d17-2e82-4f07-8ee7-c20931174da9), [`Outline details`](https://www.recraft.ai/styles/87cf0b6b-a495-4436-8598-897a71bf5db6), [`Pastel gradient`](https://www.recraft.ai/styles/9819278f-d0fe-4331-b7d4-37b0cf69150a), [`Pastel sketch`](https://www.recraft.ai/styles/149f13db-53ac-4c3e-8eb7-d5d44af04405), [`Pop art`](https://www.recraft.ai/styles/c5c3026e-66e5-412e-b87a-101512d745ac), [`Pop renaissance`](https://www.recraft.ai/styles/99c21c89-7de0-46dc-b447-e519be2c47f0), [`Street art`](https://www.recraft.ai/styles/c42616cd-f2c2-459c-a745-6f0ac86589d7), [`Tablet sketch`](https://www.recraft.ai/styles/c0c1b91e-8b46-4243-8f0b-fdea56c43701), [`Urban Glow`](https://www.recraft.ai/styles/acf7c4d0-ac61-4fb4-a3fa-d3a1fde6c014), [`Urban sketching`](https://www.recraft.ai/styles/ce069e6f-09bc-482b-9a05-fabf6658f27f), [`Young adult book`](https://www.recraft.ai/styles/e3838b40-dbac-450c-a06c-2c52b3b790c7), [`Young adult book 2`](https://www.recraft.ai/styles/b2449e5b-6954-453e-a168-c6e743dafd9a), [`Seamless Digital`](https://www.recraft.ai/styles/a054fca2-259e-449e-9b25-a7b2ec7a9aa3); * Emblem styles: [`Prestige Emblem`](https://www.recraft.ai/styles/1c2f96d5-c2cf-4311-909f-44190a8ea153), [`Pop Graphic`](https://www.recraft.ai/styles/a4681bb8-4e6b-47c6-ae61-d4dcdf9b9dbc), [`Stamp`](https://www.recraft.ai/styles/dac4f551-db80-4cdd-8778-8d6ad4fa992b), [`Punk Graphic`](https://www.recraft.ai/styles/84789d26-1153-4faf-a1d1-c3e94d7a0035), [`Vintage Emblem`](https://www.recraft.ai/styles/7faff2e0-bdce-4a2b-aea5-471f7fb88f5e); #### Recraft V3 Vector * Vector styles: [`Vector art`](https://www.recraft.ai/styles/773902ac-683a-4042-86a3-fe714eee345a), [`Line art`](https://www.recraft.ai/styles/19942671-ffa4-4867-9bbe-f5e4503dcc93), [`Linocut`](https://www.recraft.ai/styles/455a5cff-6d70-4d9a-945f-9256f279f50e), [`Color blobs`](https://www.recraft.ai/styles/7119b64b-b99d-4929-a418-3ead12ca1847), [`Engraving`](https://www.recraft.ai/styles/ce1f9161-a8dc-488f-9ff1-3f41082887b5), [`Bold stroke`](https://www.recraft.ai/styles/e33777c2-e1d3-4b16-bcdb-c7335e67a138), [`Chemistry`](https://www.recraft.ai/styles/e5aeb587-cac5-45ab-887f-bbc933a3d756), [`Colored stencil`](https://www.recraft.ai/styles/f5907d95-d032-4f70-a035-308bf8326ff3), [`Cosmics`](https://www.recraft.ai/styles/f14175ea-077f-4d8d-b1cb-e568ed906324), [`Cutout`](https://www.recraft.ai/styles/9ab9c49c-a934-4c83-af04-de6665a8d992), [`Depressive`](https://www.recraft.ai/styles/94b499fd-6408-48d1-8160-87661e93767c), [`Editorial`](https://www.recraft.ai/styles/59133ee8-2fc5-4225-bcd9-6934d21bcac8), [`Emotional flat`](https://www.recraft.ai/styles/782e2222-dbd8-46e0-bcec-1179d6c0e511), [`Marker outline`](https://www.recraft.ai/styles/1e018a0b-3299-445d-8812-06fd2f24eafa), [`Mosaic`](https://www.recraft.ai/styles/6c19cde9-52a6-40b3-9cc9-d81fe1fdd387), [`Naivector`](https://www.recraft.ai/styles/70192033-8db6-49af-b08b-f0e0043dddb8), [`Roundish flat`](https://www.recraft.ai/styles/857c6d5e-4039-45c1-8e31-12cfa25fa638), [`Segmented Colors`](https://www.recraft.ai/styles/8eafc64b-6483-46ca-858c-ef509ded10b5), [`Sharp contrast`](https://www.recraft.ai/styles/3ddad82f-d3c0-432e-ab7e-d9ed441a73c0), [`Thin`](https://www.recraft.ai/styles/f8300e4a-1b2f-4ad6-abbe-7e7180b87f30), [`Vector Photo`](https://www.recraft.ai/styles/b331f7bc-4ef6-4d17-a042-bae020cab91f), [`Vivid shapes`](https://www.recraft.ai/styles/85243d3e-dfd8-4f8b-83d7-456e72491814), [`Seamless Vector`](https://www.recraft.ai/styles/2e8f425e-b1a4-4c38-8f07-977bd8c2caf5); #### Recraft V2 * Photorealistic styles: [`Photorealism`](https://www.recraft.ai/styles/5a99ae89-3f14-433b-abb2-a1d8074b1ab1), [`Enterprise`](https://www.recraft.ai/styles/35ceb30e-50f5-46d4-9372-e95d8458a380), [`Natural light`](https://www.recraft.ai/styles/030886da-f6b9-4730-a1c1-4ce54a610a81), [`Studio photo`](https://www.recraft.ai/styles/0295270b-e050-45ad-8711-e29d1ac87d10), [`HDR`](https://www.recraft.ai/styles/07d689f1-ee16-49c4-a103-befb59bd7032), [`Hard flash`](https://www.recraft.ai/styles/4cb07726-d352-48cf-8f9b-b06eed4e0563), [`Motion blur`](https://www.recraft.ai/styles/cee8bee7-2085-4aca-91d5-2e3939f11dc8), [`Black & white`](https://www.recraft.ai/styles/6fefbfcf-9c93-4bce-8b31-bcbd86327d66), [`Product photo`](https://www.recraft.ai/styles/570c6fb3-1b01-4da0-afc9-0cd9997ffcc2); * Illustration styles: [`Illustration`](https://www.recraft.ai/styles/b8772060-8802-4f3c-b051-73349b1017a5), [`3D render`](https://www.recraft.ai/styles/4659b2d6-a580-4287-849e-96190d7dbf30), [`Glow`](https://www.recraft.ai/styles/077aeb2b-fca1-49e6-805b-6174bf55ee1a), [`Watercolor`](https://www.recraft.ai/styles/568daa6a-9631-44e7-a615-ce547146ae52), [`Hand-drawn`](https://www.recraft.ai/styles/a389a8b9-1209-463f-b816-0c08eb4f93c4), [`Kawaii`](https://www.recraft.ai/styles/f8f1bc1c-1ae7-43ea-a36f-c81049feee01), [`Grain`](https://www.recraft.ai/styles/9fe2962e-ac66-4d41-8ff4-38dbdfc79835), [`Bold Sketch`](https://www.recraft.ai/styles/41ef6ca6-8797-4cf1-87c6-522f425f2059), [`Pencil sketch`](https://www.recraft.ai/styles/f6e787d5-80cc-415e-b68d-860a0d25dafb), [`Retro Pop`](https://www.recraft.ai/styles/a28c4388-81fd-497c-a7ea-a44013636071), [`Clay`](https://www.recraft.ai/styles/c7aaa207-769f-4449-bdc4-c9d8cbcdbb09), [`Risograph`](https://www.recraft.ai/styles/902997d1-72a5-4951-81b8-042732f32d26), [`Psychedelic`](https://www.recraft.ai/styles/1f9a9d70-2d71-4b85-9106-646bbddf1fd0), [`Seamless Digital`](https://www.recraft.ai/styles/655b577f-a89f-4015-8a95-12b8daa2414b), [`Color engraving`](https://www.recraft.ai/styles/a36fc277-4f70-445b-bc1a-c50412c8c439), [`Pixel art`](https://www.recraft.ai/styles/7c8c1d97-17c8-44b4-a252-2808d44795a6), [`80's`](https://www.recraft.ai/styles/5e86f847-766c-4479-a6a0-645a54d77e06), [`Voxel art`](https://www.recraft.ai/styles/85770603-257a-4588-a9e7-73064991c1b7); #### Recraft V2 Vector * Vector styles: [`Vector art`](https://www.recraft.ai/styles/807ce9bc-7e36-4efa-a6cb-11cda1cecd19), [`Line art`](https://www.recraft.ai/styles/da8ce568-08d9-4bfb-9ef0-2984ec6391e3), [`Linocut`](https://www.recraft.ai/styles/3fc31d5e-a4ea-436c-be1a-787ec1cc76d8), [`Cartoon`](https://www.recraft.ai/styles/6cdae472-693d-478a-91fc-83eaf7c91a0c), [`Flat 2.0`](https://www.recraft.ai/styles/bcfabc0a-dec4-479b-962f-b306609e90ee), [`Color blobs`](https://www.recraft.ai/styles/2dc79001-283c-4c1e-ac01-49a95e21175c), [`Vector Kawaii`](https://www.recraft.ai/styles/6caa7c53-ad06-4d3b-827d-5bb81e782659), [`Doodle Line art`](https://www.recraft.ai/styles/1eb016d9-46d0-4b4e-bf1d-76296f43b471), [`Seamless Vector`](https://www.recraft.ai/styles/22e6d2f6-7951-49c4-aaaa-b7db63cb880a), [`Engraving`](https://www.recraft.ai/styles/100ae2cf-358a-46d2-ab13-dda7079c0460); * Icon styles: [`Icon`](https://www.recraft.ai/styles/645cb7ca-4be3-4d90-9f16-e9c26f6812f1), [`Outline`](https://www.recraft.ai/styles/5cf68b22-c903-48c5-882b-16872ef2d124), [`Pictogram`](https://www.recraft.ai/styles/be9ebb8f-312a-4d8e-b7ba-8276b05a2108), [`Colored outline`](https://www.recraft.ai/styles/7513cd33-5b71-4e04-8fed-becdd2672b4a), [`Doodle`](https://www.recraft.ai/styles/a2ba0c04-e40d-4210-8ed2-46d3537c546c), [`Colored shape`](https://www.recraft.ai/styles/1c5ca809-1bce-4b1c-8b14-79a5bc6df150), [`Gradient outline`](https://www.recraft.ai/styles/a7272add-907f-49f0-860d-d8b6d2dd0995), [`Offset doodle`](https://www.recraft.ai/styles/bdb513fc-cf7f-4e0f-be45-ef18bd693d03), [`Gradient shape`](https://www.recraft.ai/styles/65dcb522-d206-4561-bf93-9ae650f73136), [`Broken line`](https://www.recraft.ai/styles/d700cad3-94da-4204-b753-dbb75487edfc), [`Offset fill`](https://www.recraft.ai/styles/5a905231-9aee-4d4b-a0bb-eda7e79a0e95); # Swagger Source: https://www.recraft.ai/docs/api-reference/swagger OpenAPI is a standard format for describing RESTful APIs that allows those APIs to be integrated with tools for a wide variety of applications, including testing, client library generation, IDE integration, and more. To get started, check out the [Specification](https://external.api.recraft.ai/doc/#/). # Character consistency Source: https://www.recraft.ai/docs/best-practices/character-consistency Character consistency refers to maintaining a recognizable subject across multiple image generations. This can include not only people or mascots, but also products or objects—such as a specific chair, a branded package, or a hat. While Recraft doesn’t offer a dedicated character-tracking feature, you can achieve strong visual continuity using a combination of prompt engineering, style control, reference images, frames, and external model tools. ## What counts as a "character" * People (e.g., a red-haired woman in a space suit) * Mascots (e.g., a cartoon beaver used across marketing visuals) * Objects or products (e.g., the same pair of sunglasses shown in multiple lifestyle settings) ### Techniques for maintaining consistency * **Use a detailed prompt:** Start with a prompt that clearly defines your subject’s distinguishing traits. For example:\ *“Young woman with curly red hair, freckles, and an astronaut suit.”*\ This establishes a verbal template for regenerating or varying the character later.\ **Maintain a consistent style**: Apply the same style (or saved custom style) across all related images. Recraft interprets characters through the lens of visual style—keeping it consistent reduces unwanted shifts in appearance. * **Use frames to evolve or isolate specific traits**: Insert the character into a Frame to edit selected parts—such as changing clothes, expressions, or poses—without altering the full image. This is especially useful for storytelling or iterative design. * **Use reference images to preserve likeness**: Add a previously generated or uploaded image of your subject to the canvas and set it as a visual reference. Recraft will use this context to influence how new images are generated, helping preserve identity and structure. * **Use external models for remixing or variation**: You can attach one or more images (of a person, object, or mascot) to a prompt using the external model option. For example: * Combine two characters or items into a single scene. * Generate style or activity variations, such as:\ *“Make my selfie in the style of The Simpsons.”*\ *“Make me laughing.”*\ *“Make me riding a bike.”* These techniques can be layered together to iteratively build a series of images that feature a character or product with recognizable, coherent features, even as you explore different styles, scenes, or compositions. # Prompting and image generation Source: https://www.recraft.ai/docs/best-practices/prompting-and-image-generation * If the model isn’t interpreting your prompt as expected, try switching to a more general style, like **Recraft V3**, **Photorealism**, or **Illustration,** instead of a custom or niche style. * For logo designs with text that feel too rigid, enable **Negative prompt → Avoid text in prompt → Yes**. This can produce more expressive and creative text layouts, though it may introduce typos. * If prompt comprehension is poor, try lowering the **Artistic level**. This reduces creative variance and often improves adherence to prompt details. * If you're having trouble generating content that avoids the edges of the image, place the image in a **Frame** and use **Outpainting** to expand around it. * For small anatomical details (like faces or hands) that look distorted, run **Creative Upscale**. This often improves accuracy and visual quality. ### How to create mockups with designs If you’re working with a **complex surface** and need to experiment with how the design fits, such as adjusting size, placement, or angle, use **Convert to mockup**. If the mockup involves a **flat surface** where the design can be placed in a single, fixed way and doesn’t require experimenting with placement, use **prompt-based editing**.\ For example: a street banner or a phone screen. # Introduction Source: https://www.recraft.ai/docs/index Welcome to one of the world's top ranked AI image tools for designers. ## Getting started Generate your first image with Recraft. Follow these simple steps to create an image. ## Explore Recraft's tools and features Design something amazing. Get better results with Recraft's powerful editing features. Apply a curated style or learn how to create your own to stay on-brand. Create realisitc product mockups in minutes. Set up your own color swatches and palettes. ## Learn about plans, ownership, and getting help Everything you need for a meaningful experience. Learn what's included with our paid plans. Understand your commercial rights and copyright for generated content. Contact support, report a bug, and request a feature. View most frequently asked questions. ## Lookign for inspiration? Browse our community gallery to see what's possible with Recraft. # Getting started Source: https://www.recraft.ai/docs/mcp-reference/getting-started Recraft supports the [Model Context Protocol](https://modelcontextprotocol.io/) (MCP), enabling AI agents such as Claude, Cursor, and other MCP-compatible clients to interact with Recraft’s image generation and editing tools through a standardized interface. By connecting Recraft's MCP server to your client, you can generate and edit high-quality raster and vector images directly from your agent environment without switching tools. Recraft offers two ways to connect: * **Remote MCP server** — a hosted server at `https://mcp.recraft.ai/mcp` that requires no local installation. Authenticate with your API key and connect directly from any MCP-compatible client. See [Remote MCP server](./remote-server) for setup instructions. * **Local MCP server** — a self-hosted server you run on your machine. Useful if you want to store generated images locally or need full control over the setup. See [Setup](./setup) for installation options. ## Supported operations Recraft’s MCP server exposes the following operations: * Raster and vector image generation * Raster and vector image editing * Creating and using custom styles * Vectorization of raster images * Background removal and replacement * Upscaling of raster images For the source code, visit [Recraft MCP GitHub](https://github.com/recraft-ai/mcp-recraft-server). # Local MCP Server Source: https://www.recraft.ai/docs/mcp-reference/local-server ## Prerequisites 1. **Recraft API key** * Access it in the [API section](https://www.recraft.ai/profile/api) of your account **Profile**. * Note: You will need to buy API units (credits) before you can generate an API key. 2. **MCP client installed** * For example, [Claude Desktop](https://claude.ai). ## Option 1: Claude Desktop Extensions You can set up the Recraft MCP server in Claude using [Claude Desktop Extensions](https://www.anthropic.com/engineering/desktop-extensions). 1. Download **mcp-recraft-server.dxt** from the [latest release](https://github.com/recraft-ai/mcp-recraft-server/releases/latest/download/mcp-recraft-server.dxt). 2. Double-click the file to open it with Claude Desktop Extensions. 3. In Claude Desktop, click **Install**. 4. Fill out the setup form: * Paste your Recraft API key obtained from your [Profile API page](https://www.recraft.ai/profile/api). * (Optional) Specify a local path for generated image storage, or indicate that all results should be stored remotely. 5. Enable the server. If you encounter installation issues, make sure you have the latest version of Claude Desktop installed. ## Option 2: Smithery 1. Find this MCP server on [Smithery](https://smithery.ai/server/@recraft-ai/mcp-recraft-server). 2. Install from Smithery. * Note: all generation results will be stored remotely. * If you want to store results locally, use Claude Desktop Extensions or manual setup instead. ## Option 3: Manual setup ### Requirements * Ensure **Node.js** is installed on your machine. * If not, install from [nodejs.org](https://nodejs.org). * You will need to run `npx` or `node` commands in your terminal. ### From NPM 1. Open your **Claude Desktop configuration file**: `claude_desktop_config.json`. 2. Modify the file to add the following configuration snippet (replace with your API key and desired settings): ```json theme={null} { "mcpServers": { "recraft": { "command": "npx", "args": [ "-y", "@recraft-ai/mcp-recraft-server@latest" ], "env": { "RECRAFT_API_KEY": "", "IMAGE_STORAGE_DIRECTORY": "", "RECRAFT_REMOTE_RESULTS_STORAGE": "" } } } } ``` ## Manual setup To set up the Recraft MCP server manually, you will need **Node.js** installed so you can run `npx` or `node` commands in your terminal. If you don’t already have Node.js, download it from [nodejs.org](https://nodejs.org/en/download). ### From NPM 1. Open your `claude_desktop_config.json` file. 2. Add the following configuration block, replacing the placeholders with your details: ```json theme={null} { "mcpServers": { "recraft": { "command": "npx", "args": [ "-y", "@recraft-ai/mcp-recraft-server@latest" ], "env": { "RECRAFT_API_KEY": "", "IMAGE_STORAGE_DIRECTORY": "", "RECRAFT_REMOTE_RESULTS_STORAGE": "" } } } } ``` ### From source 1. Clone the repository: ```bash theme={null} git clone https://github.com/recraft-ai/mcp-recraft-server.git ``` 2. Navigate into the cloned directory and build the project: ```bash theme={null} npm install npm run build ``` 3. Modify your `claude_desktop_config.json` file with the following configuration: ```json theme={null} { "mcpServers": { "recraft": { "command": "node", "args": [ "/dist/index.js" ], "env": { "RECRAFT_API_KEY": "", "IMAGE_STORAGE_DIRECTORY": "", "RECRAFT_REMOTE_RESULTS_STORAGE": "" } } } } ``` ### Environment variables You can configure the following parameters in the `env` section: * **`RECRAFT_API_KEY`** *(required)* Your [Recraft API key](https://www.recraft.ai/profile/api). * **`IMAGE_STORAGE_DIRECTORY`** *(optional)* Local directory to store generated images. Defaults to: ``` $HOME_DIR/.mcp-recraft-server ``` This value is ignored if `RECRAFT_REMOTE_RESULTS_STORAGE="1"`. * **`RECRAFT_REMOTE_RESULTS_STORAGE`** *(optional)* Set to `"1"` to store all generated images remotely. In this case, only URLs are returned and `IMAGE_STORAGE_DIRECTORY` is ignored. # Remote MCP server Source: https://www.recraft.ai/docs/mcp-reference/remote-server Connect to Recraft's hosted MCP server without any local installation Recraft provides a hosted remote MCP server at `https://mcp.recraft.ai/mcp`. It supports the Streamable HTTP transport and uses OAuth 2.0 for authentication. The remote server exposes all functionality available through the [Recraft API](https://www.recraft.ai/docs/api-reference/endpoints), including image generation, editing, vectorization, upscaling, background removal and replacement, and custom style creation. ## Authentication The remote MCP server uses OAuth 2.0. When you connect for the first time, your MCP client will initiate an OAuth flow and redirect you to Recraft to authorize access. Once authorized, the client stores the token and handles authentication automatically on subsequent requests. ## Setup by client ### Claude Code Run the following command in your terminal: ```bash theme={null} claude mcp add --transport http recraft https://mcp.recraft.ai/mcp ``` Then run `/mcp` in a session to authenticate via OAuth. ### Claude Desktop Open your `claude_desktop_config.json` file and add the following configuration: ```json theme={null} { "mcpServers": { "recraft": { "command": "npx", "args": [ "-y", "mcp-remote", "https://mcp.recraft.ai/mcp" ] } } } ``` On first use, `mcp-remote` will open a browser window to complete the OAuth flow. ### Cursor Open your `~/.cursor/mcp.json` file and add the following configuration: ```json theme={null} { "mcpServers": { "recraft": { "url": "https://mcp.recraft.ai/mcp" } } } ``` ### Visual Studio Code Open your `.vscode/mcp.json` file (or your user settings) and add: ```json theme={null} { "servers": { "recraft": { "type": "http", "url": "https://mcp.recraft.ai/mcp" } } } ``` ### Other MCP clients Any MCP-compatible client that supports Streamable HTTP transport and OAuth can connect using: * **URL:** `https://mcp.recraft.ai/mcp` * **Transport:** Streamable HTTP If your client does not support remote OAuth flows directly and requires a stdio-based command, use [`mcp-remote`](https://www.npmjs.com/package/mcp-remote): ```bash theme={null} npx -y mcp-remote https://mcp.recraft.ai/mcp ``` ## Supported operations The remote server supports the full set of image operations available through the Recraft API, including: * Generate raster and vector images from text prompts * Create image variations and image-to-image transformations * Edit selected parts of an image * Remove selected objects or regions * Remove, replace, or generate backgrounds * Convert raster images into vector graphics * Upscale images for either sharper or more detailed results * Create custom styles from reference images * View account details such as your remaining credits # Tools Source: https://www.recraft.ai/docs/mcp-reference/tools Recraft's MCP server allows use of the following tools: | Tool name | Description | Parameters | Price | | -------------------- | ----------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------- | --------------------------------------- | | `generate_image` | Generates raster/vector images from a prompt | - `prompt`
- `style`
- `size`
- `model`
- `number of images` | \$0.04 / \$0.08 per raster/vector image | | `create_style` | Creates a style from a list of images | - `list of images`
- `basic style` | \$0.04 | | `vectorize_image` | Vectorizes a raster image into a vector format | - `image` | \$0.01 | | `image_to_image` | Generates raster/vector images from an image + prompt | - `image`
- `prompt`
- `similarity strength`
- `style`
- `size`
- `model`
- `number of images` | \$0.04 / \$0.08 per raster/vector image | | `remove_background` | Removes the background of an image | - `image` | \$0.01 | | `replace_background` | Generates a new background in an image from a prompt | - `image`
- `prompt for background`
- `style`
- `size`
- `model`
- `number of images` | \$0.04 / \$0.08 per raster/vector image | | `crisp_upscale` | Upscales image resolution without altering content | - `image` | \$0.004 | | `creative_upscale` | Upscales image resolution while regenerating details | - `image` | \$0.25 | | `get_user` | Retrieves information about the user and balance | – | – | For a detailed explanation of each tool, its parameters, and pricing, visit the [Recraft API docs](https://www.recraft.ai/docs/api-reference/getting-started). # Billing and invoices Source: https://www.recraft.ai/docs/plans-and-billing/billing Subscriptions can be billed on a **monthly** or **annual** basis. Annual plans include a discount compared to monthly billing. **Accepted payment methods** include credit cards, debit cards, and bank transfers. ## Canceling your subscription You can cancel your subscription at any time through the **Manage subscription** option in your profile. Your plan will remain active until the end of the current billing cycle. If you have used fewer than 30 credits over the lifetime of your account and your most recent payment was made within the past 30 days, you may request a refund. To do so, cancel your subscription and contact Recraft support at [help@recraft.ai](mailto:help@recraft.ai). Refunds are issued to the original payment method. ## Switching your billing cycle from yearly to monthly To change from yearly to monthly billing, you will need to first cancel your yearly subscription, then start a new subscription on a monthly billing cycle: 1. At the bottom left of the page, click on your name, then select **Manage subscription**. 2. Click **Cancel subscription** at the bottom, then complete the cancellation flow. 3. Go to the [Pricing page](https://www.recraft.ai/pricing) and click the toggle to select monthly billing. 4. Complete the purchase flow. Your yearly subscription will remain active until the end of your current billing period. After that, your plan will renew on a monthly basis. ## How to download an invoice or receipt 1. For personal accounts: Click on your profile avatar in the upper right, then select **Manage subscription**.\ For teams accounts: Open the **Workspace** tab, then select **Manage subscription**. 2. Click **Manage billing details**. 3. In the Invoice history section at the bottom, click on your desired invoice. 4. Choose from **Download invoice** or **Download receipt**. # Commercial rights and ownership Source: https://www.recraft.ai/docs/plans-and-billing/commercial-rights-and-ownership Recraft’s usage rights depend on your subscription plan. Images generated under the **Free plan** are publicly visible in the community gallery and remain the property of Recraft. These images are not licensed for commercial use. For example, they typically cannot be listed on stock platforms, which require proof of ownership. Images generated under a **paid plan** grant you full ownership and commercial rights. You may use these assets for any purpose, including marketing, branding, product packaging, or other professional applications. Paid plan images remain private and under your ownership even after your subscription ends; they are not made public. Note: Regardless of plan, assets generated with Recraft may not be used to train any artificial intelligence models, systems, networks, or similar technologies. Learn more about image rights and ownership on [the Recraft blog](https://www.recraft.ai/blog/ownership-and-commercial-use-faq). # Credits Source: https://www.recraft.ai/docs/plans-and-billing/credits Recraft uses credits to power all image generation and editing operations. Free users receive 30 credits per day. Paid plans include a monthly credit allowance, which resets at the start of each billing cycle. Credits are consumed based on the type of action performed. The table below summarizes typical credit costs: | Action | Credits | | :-------------------------- | :------ | | Convert image to mockup | 2 | | Creative upscale | 20 | | Crisp upscale | 1 | | Erase region | 1 | | Generate a mockup | 2 | | Generate a raster image | 1 | | Generate a vector image | 2 | | Modify region (raster) | 1 | | Modify region (vector) | 2 | | Remove background (raster) | 1 | | Remove background (vector) | 2 | | Replace background (raster) | 1 | | Replace background (vector) | 2 | | Vectorization | 1 | ## Credit types and expiration * Subscription credits reset monthly (or daily for Free users) and do not roll over. * Top-up credits never expire and remain available until used, even if the subscription ends. * When subscription credits are depleted, top-up credits are used automatically. * Credits can be earned through the referral program or purchased as needed (top-ups are available to paid users only). ## Credit top-ups Top-up credits can be purchased at any time, even before subscription credits are fully used. This ensures uninterrupted access to generation tools. It increments by 200, 800, 1000, and 1600 credits. They never expire and are used after your monthly credits. and are only available to users on paid plans. **How to top up credits**: 1. Open Recraft and sign in. 2. Click on the credit counter icon in the upper right corner, then click **Buy more**. 3. A dialog box will appears to confirm your purchase, click **Add credits** to confirm. 4. Once payment is processed, the updated credit balance will appear (reload the page if necessary). Some models are provided by external providers and may have different credit costs. See the [External Models](https://www.recraft.ai/docs/recraft-studio/image-generation/external-models) page for more details. If assistance is needed, contact support at [help@recraft.ai](mailto:help@recraft.ai). Notes: * Top-up packs are available in fixed increments of 200, 800, 1000, and 1600. * There is no limit to the number of top-ups that can be purchased. * Top-up credits are preserved across billing periods and subscriptions. # Deleting an account Source: https://www.recraft.ai/docs/plans-and-billing/deleting-an-account Deleting your account permanently removes your access to Recraft. According to our pricing policy, refunds are not available for deleted accounts. ### Desktop app 1. Log in to your account. 2. Click the **Profile** icon in the top-right corner to open the menu. 3. Select **Profile**, then choose **Delete account**. ### Mobile app 1. Tap your **Profile** icon in the top-right corner. 2. Select **Delete account** from the menu. # Free plan Source: https://www.recraft.ai/docs/plans-and-billing/free-plan The Free plan is ideal for trying out Recraft’s core functionality without a subscription. It offers a daily credit allocation of 30 credits, allowing you to explore Recraft’s image and vector generation tools along with basic editing features. You can upload up to three images per day, and each prompt produces up to two images at a time. Credits reset every 24 hours. Credit top-ups are only available on paid plans. Generated images in the Free plan are public and may appear in the community gallery. # Paid plans Source: https://www.recraft.ai/docs/plans-and-billing/paid-plans [***Refer to the Recraft pricing table for the latest credit allowances and subscription rates.***](https://www.recraft.ai/pricing) Recraft offers paid subscriptions for individual creators and teams with both monthly and annual billing options (annual billing provides up to 20% discount). For individual users, the **Basic plan** starts at: * **\$12 per month (billed monthly)** for 1,000 credits * **\$10 per month (billed annually, \$120 per year)** for 1,000 credits The **Pro plan** starts at: * **2000 credits** — \$20 per month (monthly) or \$16 per month (billed annually) * **4,000 credits** — \$40 per month (monthly) or \$32 per month (billed annually) * **8,000 credits** — \$80 per month (monthly) or \$64 per month (billed annually) * **16000 credits** — \$160 per month (monthly) or \$128 per month (billed annually) Credits renew monthly. Additionally, you can purchase top-up credits, which never expire and remain available until used, even if the subscription ends. For collaborative use, the **Teams plan** is available at: * **2000 credits** — \$22 per month (monthly) or \$18 per month (billed annually) * **4,000 credits** — \$44 per month (monthly) or \$35 per month (billed annually) * **8,000 credits** — \$70 per month (monthly) or \$88 per month (billed annually) * **16000 credits** — \$176 per month (monthly) or \$141 per month (billed annually) The Teams plan includes: * Shared custom styles * Centralized account management * Premium 24/7 support * Single Sign-On (SSO) All paid plans include private image generation and commercial usage rights while subscribed. ## Premium plans The Premium plans are intended for individuals who want to generate private images and retain [full ownership](/plans-and-billing/commercial-rights-and-ownership). They include a monthly credit allowance and unlock access to all core editing and generation tools, as well as the ability to top up with additional credits as needed. Check out the full details on our [Pricing Page](https://recraft.ai/pricing) to see everything included! ## Credit top-ups You can also top up credit packs at any time, in increments of 200, 800, 1000 and 1600. They never expire and are used after your monthly credits. Learn more about [credits](/plans-and-billing/credits). ## Teams Recraft offers paid plans for teams and enterprise organizations. To learn more, visit the [Teams page](/teams/teams-overview). # Upgrading and downgrading a plan Source: https://www.recraft.ai/docs/plans-and-billing/upgrading-and-downgrading-a-plan ## Upgrading Once confirmed, your plan will be upgraded immediately. Your billing will be prorated based on the remaining time on your current plan. **To upgrade to a paid Pro plan from a Free plan**: 1. Click **Upgrade** in the upper right screen. 2. Click **Get started** in the **Pro** plan. 3. Select the amount of credits you'd like to purchase, then click **Continue to payment**. 4. In the Stripe checkout window, confirm your new plan and complete the purchase. **To increase the credits on your current Pro plan**: 1. Open the **Profile menu** in the upper-right corner of the app. 2. Click **Pricing and plans**. 3. On the pricing page, locate the Pro plan in the center, then click **Current plan**. 4. Select the amount of credits you'd like to purchase then click **Continue to payment**. 5. In the Stripe checkout window, confirm your new plan and complete the purchase. ## Downgrading **To downgrade to the Free plan from a Pro plan**: 1. Open **Profile** in the upper-right corner of the app. 2. Click **Manage subscription**. In the Subscription box, click **Change plan**. 3. Click **Downgrade** under **Free**. 4. Click **Cancel subscription**. 5. Confirm changes. **To downgrade the credit limit from a Pro plan**: 1. Open **Profile** in the upper-right corner of the app. 2. Click **Manage subscription**. In the Subscription box, click **Change plan.** 3. Click **Current plan** under **Pro**. 4. Select the amount of credits you'd like to purchase, then click **Continue to payment**. 5. Confirm changes. After downgrading, you may receive a prorated amount of unused credits on your Stripe account that can be applied to future purchases. # Attributes Source: https://www.recraft.ai/docs/prompt-engineering-guide/artistic-principles/attributes Attributes shape the fine details of the image. They determine lighting behavior, color relationships, texture, materials, and typographic structure, adding polish without altering the core composition. Once the subject, composition, medium, and style are defined, attributes refine the surface qualities and atmosphere of the image. These cues help the model interpret nuance and realism while keeping the overall structure intact. **Lighting** Influences mood and focus, for example soft morning light, dramatic side lighting, diffuse overcast light, or harsh fluorescent light. **Color and palette** Shapes mood and visual identity, for example muted earth tones, high-contrast palette, pastel gradient, or bright complementary colors. **Texture and materials** Describe how surfaces interact with light such as matte metal, polished glass, rough linen, or grainy paper. **Typography and layout** (for graphic or vector work) Specify typographic hierarchy and structure such as bold sans-serif headline, subtle serif caption, minimal grid layout, or clean centered title. ## Examples
"*Portrait of a woman in a studio*" "*Portrait of a woman in studio with soft window light from left, gentle shadows, warm golden hour glow, natural diffused illumination*" "*Close-up portrait of woman in studio, face and shoulders filling frame, soft window light from left creating gentle shadows across features, muted earth tone palette, matte skin texture with natural detail, shallow depth of field f/2.8 with background softly blurred, slight film grain adding warmth, golden atmospheric glow, gentle rim lighting highlighting hair edges, natural elegant expression, contemporary editorial style, intimate framing, 85mm lens perspective*" # Medium Source: https://www.recraft.ai/docs/prompt-engineering-guide/artistic-principles/medium The medium establishes the visual language of the image. It defines the format or technique the model should follow and shapes how edges, textures, lighting, and materials are rendered. Specify the medium clearly, whether it’s a photograph, watercolor painting, 3D render, pencil sketch, or flat vector illustration. This helps the model understand the visual framework it should use when generating the image. ## Examples
"*Single whole red apple centered in frame, medium size filling one-third of image, natural window light, realistic details with water droplets, professional food photography, clean white background*" "*Single whole red apple centered in frame, medium size filling one-third of image, watercolor painting, soft flowing edges, translucent washes, vibrant red and green palette, visible paper texture, white background*" "*Single whole red apple centered in frame, medium size filling one-third of image, flat vector illustration, simple geometric shape, clean minimal design, solid red color with highlight, modern graphic style, white background*" # Style Source: https://www.recraft.ai/docs/prompt-engineering-guide/artistic-principles/style Style directs the model toward a specific artistic or visual tradition. It keeps the image coherent by aligning it with a clear aesthetic direction. Style can reference artistic movements such as Impressionism or Bauhaus, genres like editorial fashion photography or minimal poster design, or techniques like collage, ink sketch, or isometric render. Clear stylistic cues help the model stay focused on a consistent look rather than blending unrelated influences. ## Examples
"*Single whole red apple centered in frame, medium size filling one-third of image, authentic 1950s vintage advertisement style with aged paper texture, bold oversaturated red and green colors typical of mid-century print, prominent halftone dot screen pattern visible throughout, slight faded edges and subtle grain, retro commercial appeal with nostalgic quality, printed ephemera aesthetic, slight color separation misalignment for authentic vintage printing effect, classic Americana advertising look, warm sepia undertones, old magazine ad feeling*" "*Single whole red apple centered in frame, medium size filling one-third of image, Japanese minimalist aesthetic, zen simplicity, muted natural red tone, wabi-sabi philosophy, clean white empty space*" "*Single whole red apple centered in frame, medium size filling one-third of image, organic surrealist transformation with skin seamlessly morphing into liquid mercury on one side, roots and branches emerging from bottom creating impossible life cycle, butterflies made of apple flesh flying away leaving hollow spaces, soft pastel dreamlike background with melting clock elements à la Dalí, fruit existing in multiple states of matter simultaneously, poetic visual metaphor style, contemporary surrealist fine art photography, impossible beauty*" # Vibe Source: https://www.recraft.ai/docs/prompt-engineering-guide/artistic-principles/vibe Vibe sets the emotional quality of the image. It shapes mood and atmosphere so the output reflects your intended tone. Choose descriptors that clearly express the feeling you want to convey. Terms like serene, dramatic, playful, or tense help the model establish a consistent emotional direction. ## Examples
"*Single whole red apple centered in frame, medium size filling one-third of image, fresh morning harvest mood, crisp and inviting, bright natural light, healthy vibrant feeling, clean white background*" "*Single whole red apple centered in frame, medium size filling one-third of image, nostalgic autumn memory, warm golden tone, rustic countryside feeling, peaceful contemplative mood, soft neutral background*" "*Single whole red apple centered in frame, medium size filling one-third of image, dramatic moody atmosphere, dark black background with spotlight, intense shadows, theatrical chiaroscuro lighting, fine art still life*" # Closing notes Source: https://www.recraft.ai/docs/prompt-engineering-guide/closing-notes Prompting is a creative practice. The more you experiment, observe, and refine, the more natural it becomes to guide an image model with confidence. Each attempt teaches you something about how language shapes the result, revealing what matters most in your description and where a small adjustment can shift the entire mood or structure of an image. Start with what you see in your mind, describe it simply, and build from there. Some prompts will work immediately. Others will take a few iterations. Both are part of the process. Over time, you'll start to recognize patterns, develop preferences, and form a personal vocabulary that makes prompting feel intuitive rather than experimental. # Composition Source: https://www.recraft.ai/docs/prompt-engineering-guide/core-principles/composition Composition defines how the subject is positioned and viewed within the frame. Clear spatial instructions help the model understand perspective, balance, and overall framing. Specify how the subject should appear by describing camera angle, crop, pose, and perspective. A structured description helps the model interpret spatial relationships and produce a coherent image. ## Examples
"*Hands holding ceramic mug centered in frame, direct frontal view, cozy sweater sleeves visible, neutral background, soft even lighting, lifestyle close-up*" "*Hand resting beside coffee cup on left third of frame, full cup of coffee with steam on right side, wooden table surface, morning natural light, editorial beverage photography, minimalist composition*" "*Extreme close-up of hands kneading dough, tight crop on fingers and flour, texture details visible, warm directional light, macro detail shot*" # Context Source: https://www.recraft.ai/docs/prompt-engineering-guide/core-principles/context Context adds story, environment, and atmosphere. It helps the model understand the setting and the conditions that surround the subject. Provide supporting details such as environment, style, mood, or lighting to help anchor the scene. You do not need long descriptions, only enough to guide the model. Too little context leads to unpredictable results, while too much can restrict the outcome. Aim for a balanced level of detail that supports your intention without over-constraining the image. ## Examples
"*Full body shot of humanoid robot standing centered in frame, complete figure visible from head to feet, neutral white studio background, soft even lighting, clean product photography, frontal view, sleek modern design*" "*Full body shot of humanoid robot standing centered on foggy city street at night, complete figure visible, neon signs reflecting on metallic surface, rain-wet pavement around feet, cyberpunk atmosphere, cinematic photography, moody blue and pink tones, futuristic urban setting*" "*Full body shot of humanoid robot standing centered in overgrown abandoned greenhouse, complete figure visible from head to feet, dramatic contrast between advanced technology and wild nature reclaiming space, vines growing around base, soft natural light through broken glass roof, post-apocalyptic meets botanical, cinematic editorial photography*" # Subject Source: https://www.recraft.ai/docs/prompt-engineering-guide/core-principles/subject Start your prompt with the subject. This helps the model identify what the viewer should notice first and anchors the image around a clear focal point. Describe the primary character, object, or setting in simple, descriptive terms. Stating the subject clearly guides the model’s interpretation and keeps the output aligned with your intention. ## Examples Basic chair A wooden chair A detailed image of a wooden chair
"*A chair*" "*Wooden chair*" "*Mid-century modern walnut chair positioned beside tall window with afternoon sunlight casting diagonal shadows across seat, minimalist Scandinavian interior with white walls and light oak flooring, architectural photography style, clean composition with negative space*" # Level of detail Source: https://www.recraft.ai/docs/prompt-engineering-guide/depth-and-control/level-of-detail Detail level controls how much creative freedom the model has. By adjusting the amount of instruction you provide, you decide whether the model should explore broadly or render with precision. The level of detail in a prompt determines how much guidance the model receives and how closely it will follow your intention. Choosing the right amount of detail helps you balance predictability with creative exploration. Use level of detail intentionally. Expand it when you need specific results, and reduce it when you want variation and creative freedom. ## Examples
"*Avant-garde fashion portrait, sculptural headpiece*" "*Close-up avant-garde fashion portrait with dramatic white sculptural paper headpiece, model's face and shoulders visible, muted sage green background, soft studio lighting, editorial style*" "*Close-up editorial fashion portrait of model with natural beauty and no makeup wearing dramatic avant-garde white paper sculptural headpiece with geometric folded elements radiating outward, complete head and headpiece fully visible within frame from top to shoulders, face centered with breathing room around edges, natural skin with visible freckles and texture, serene calm expression with eyes looking directly at camera, wearing simple white high-neck garment, muted sage green studio background creating color harmony, soft diffused lighting from front with subtle shadows defining natural facial features, contemporary high-fashion meets authentic beauty aesthetic, architectural fashion photography, shot on 85mm lens at f/2.8 with gentle background blur, entire sculptural element captured without cropping, muted pastel palette emphasizing natural skin tones whites and soft greens, clean balanced composition, professional editorial photography celebrating raw natural beauty*" # Prompting for image generation Source: https://www.recraft.ai/docs/prompt-engineering-guide/introduction Cover image with several Recraft AI-generated images AI image generation allows you to create visuals using natural language instead of manual tools. With just a brief description, the model interprets your prompt and generates an image based on that input. For many users, the challenge is not producing an image, it is producing the right one. Even when your idea feels clear, the output can still look unexpected or misaligned with your intention. Effective prompting is not about memorizing magic keywords. It is about understanding how models interpret language and how simple adjustments to clarity and structure can guide the results. This guide introduces Recraft’s recommended prompting approach, built around a set of core components that help you communicate your creative idea more precisely. The **core principles** cover what every strong prompt needs at a minimum. The **artistic principles** are optional layers that enhance creativity, visual richness, and stylistic identity. Once these elements are clear, you can choose the image style you want to work in and decide whether your prompt should be brief or highly detailed. # Universal prompt template Source: https://www.recraft.ai/docs/prompt-engineering-guide/prompt-templates/universal This template brings together all prompt elements in a single structured format. It shows how subject, composition, context, medium, style, vibe, and attributes work together to guide the model toward a high-quality, coherent result. ## Template `[SUBJECT + ACTION], [COMPOSITION], [CONTEXT], [MEDIUM], [STYLE], [VIBE], [ATTRIBUTES: lighting, color, texture, technical details]` Prompt: A humanoid robot with black body standing in a vast wildflower meadow with arms raised upward, captured from behind, dramatic mountain ranges stretching across the horizon in soft focus, endless field of blue and white flowers surrounding the robot, professional cinematic photograph, contemporary sci-fi style, contemplative and mystical mood, soft diffused daylight with gentle atmospheric haze, cool color palette with pastel blues, whites, and muted greens, sharp focus on the robotic silhouette, slight depth of field blur on distant mountains, shot with 35mm lens at f/2.8 “A humanoid robot with black body standing in a vast wildflower meadow with arms raised upward, captured from behind, dramatic mountain ranges stretching across the horizon in soft focus, endless field of blue and white flowers surrounding the robot, professional cinematic photograph, contemporary sci-fi style, contemplative and mystical mood, soft diffused daylight with gentle atmospheric haze, cool color palette with pastel blues, whites, and muted greens, sharp focus on the robotic silhouette, slight depth of field blur on distant mountains, shot with 35mm lens at f/2.8” ## Breakdown 1. **Subject:** “A humanoid robot with black body standing in a vast wildflower meadow with arms raised upward” 2. **Composition:** “Captured from behind, dramatic mountain ranges stretching across the horizon in soft focus” 3. **Context:** “Endless field of blue and white flowers surrounding the robot” 4. **Medium:** “Professional cinematic photograph” 5. **Style:** “Contemporary sci-fi style” 6. **Vibe:** “Contemplative and mystical mood” 7. **Attributes:** “Soft diffused daylight with gentle atmospheric haze, cool color palette with pastel blues, whites, and muted greens, sharp focus on the robotic silhouette, slight depth of field blur on distant mountains, shot with 35mm lens at f/2.8” # Prompting with Recraft V4 Source: https://www.recraft.ai/docs/prompt-engineering-guide/prompting-with-recraft-v4 Image The model is designed to operate across different levels of control while consistently delivering outputs aligned with design-level quality and visual intent. You can describe intent briefly or define constraints in detail — both approaches are valid and produce stable results. * V4 accurately follows long, highly detailed prompts with multiple constraints. * V4 produces coherent, usable results from very short prompts. * V4 performs reliably in vector illustration and logo systems. * V4 generates stylized illustrations with clear silhouette, depth, and visual hierarchy. * V4 handles 3D visuals with consistent volume and material definition. * V4 works with text predictably, including multi-language text. * V4 delivers refined, designer-level photorealism when realism is required. In some cases, a short prompt is sufficient to explore form, mood, or composition. In others, longer prompts allow you to define structure, style systems, typography behavior, or production constraints.This guide focuses on practical usage: how prompt length, structure, and level of detail affect output — and how to choose the appropriate approach for different design tasks. ## Short Prompts → Interpretive Mode Recraft is capable of making informed aesthetic decisions when provided with minimal input. **Examples** Fashion couple portrait, close up. Fashion portrait, close up. Fashion man portrait, close up.
"*Fashion couple portrait, close up.*" "*Fashion portrait, close up.*" "Fashion man portrait, close up." ## Why this works: *Even minimal prompts produce:* * Balanced framing * Controlled lighting * Clean composition * Cohesive styling Recraft fills in missing structure using internal design logic. *When to use short prompts:* * Early-stage exploration * Mood discovery * Concept sketching * Allowing the model to introduce variation Short prompts activate interpretive behavior. ## Structured Prompts → Architectural Control If you want precision, define the visual system. **Examples** Centered close-up portrait of white woman. Artistic close-up portrait of a young Caucasian woman. Ultra-photorealistic high-angle bust portrait of an adult figure.
"*Centered close-up portrait of white woman, slightly closer than waist-up, natural three-quarter turn (semi-profile), wearing a structured tailored jacket with visible seam construction and precise cut lines. Matte fabric with realistic textile texture and subtle depth. Very natural-looking model with minimal to no makeup, authentic skin texture, soft imperfections visible.Clean seamless background without a horizon line. Natural daylight, soft and diffused, coming from the side (window light feel), gentle falloff across the face, delicate shadow under the chin and along the jawline. Calm, contemporary editorial mood. Balanced negative space, intimate framing, spatial clarity, quiet realism.*" "*Artistic close-up portrait of a young Caucasian woman and a pale horse against a dark desaturated blue-grey background with subtle vignetting. The woman stands on the right in sharp head-and-torso focus, slightly angled left, looking directly into the camera with a serious, neutral expression. Fair freckled skin, light reddish-brown hair pulled back, light eyes, no jewelry. She wears a high-neck metallic gold sequined garment over a dark brown layer, reflecting soft golden highlights.In the left foreground, slightly behind her, the pale horse’s head is softly out of focus but clear in form — white coat with faint speckles, light mane, visible eye, no tack. Strong natural light from the upper left creates sharp diagonal shadows and warm highlights against the cool background. Shallow depth of field, high contrast, eye-level composition, asymmetrical and intense mood.*" "Ultra-photorealistic high-angle bust portrait of an adult figure with dark skin visible at the neck and jawline, natural pores and realistic texture. Head slightly tilted downward, face fully concealed by a lightweight reddish-orange cotton head covering with visible weave, natural folds, and small light-blue and off-white floral print. A metallic silver crown of thorns wraps around the head, pressing into the fabric with realistic tension.Silver bead strands cross the forehead horizontally and hang vertically over the fabric and chest, each bead showing accurate reflections and subtle imperfections. The figure wears a deep navy corduroy kimono-style robe with ribbed texture and visible seams, with a cream underlayer at the neckline. Clear sky background fading from deep blue to pale haze over an arid landscape. Strong natural sunlight from the front-right, sharp shadows, realistic metal reflections, high dynamic range. High-end fashion editorial style, 8K clarity, solemn and contemplative mood." **Prompt structure (from global to local):** 1. Core concept — subject(s) and scene (who and what is in the image) 2. Background and environment (where the subjects exist) 3. Primary subject framing and pose (pose and expression) 4. Physical attributes and identity details (identity and appearance) 5. Secondary subjects and spatial relationships (if needed) 6. Lighting direction and behavior 7. Camera, depth, and contrast (how the scene is captured) 8. Mood and compositional resolution **Key takeaway:**\ Structured prompts don’t make results “better.”They make outcomes intentional, controllable, and repeatable. \ **Practical tip:**\ If the image must match a specific art direction → structure the prompt.If you’re exploring → keep it minimal. Image ## Vector & Logo Design Strength Recraft V4 is unusually strong at flat graphic logic. **Examples** Minimal playful logo. Minimal playful logo. Create a collection.
"*Minimal playful logo on a deep muted green background with warm off-white or cream elements. Flat colors only — no gradients, shadows, or texture. Designed for a slow, nature-focused creative space such as a studio, independent publishing project, or mindfulness platform. Calm, grounded, human, slightly whimsical — never commercial or trendy.Bold hand-drawn chunky wordmark with soft uneven strokes, rounded edges, imperfect spacing, and slightly irregular proportions. Letters feel brush-painted or cut from paper, earthy and intuitive rather than precise. Simple plant-inspired symbols (abstract leaves, berries, seeds) are integrated into or around the lettering — flat, bold, highly simplified shapes with no outlines or fine details.Centered, compact composition forming one unified silhouette with strong legibility. Warm, natural, quietly confident identity. Avoid corporate aesthetics, sharp geometry, thin lines, decorative fonts, or tech styling.*" "*Clean modern illustration icon set featuring exactly 4 ultra-bold surreal character pictograms in a playful contemporary mascot style. Exaggerated geometric bodies, elastic limbs, dynamic confident poses, slightly asymmetrical forms, smooth hand-drawn outlines with a subtle wobble. Minimal facial features — small oval eyes, calm neutral expressions. Personality expressed through posture. Strict two-tone palette only: deep saturated crimson for all characters and graphic elements on a soft warm cream background. Flat vector aesthetic. No gradients, no shading, no texture, no shadows. Include exactly four characters: – A tall melting heart with long flexible legs walking forward – A floating crescent-moon-headed figure hugging a large abstract flower – A stretched dachshund-like creature with an oversized bow around its body – A tall matchstick-headed character with a tiny flame-shaped head holding a large geometric love letter. Add minimal decorative elements in the same crimson line style: small starbursts, abstract leaves, curved motion lines, and floating petals. Clean, balanced, poster-style composition with strong negative space. No text. No clutter. Bold, graphic, collectible art-print aesthetic.*" "*Create a collection of 12 clean vector tourism character icons arranged in a grid with four per row (three rows total). Ultra-minimal bold silhouette style, pure black solid shapes on a white background. No gradients, no textures, no outlines — only simple geometric filled forms with strong silhouette clarity. All characters must have identical tiny dot eyes, same size and placement, with no additional facial details.The icons should feel playful, abstract, slightly absurd, and clearly tourism-inspired in a modern Scandinavian-brutalist pictogram style.Include exactly these 12 characters: a mountain with minimal sunrise rays, a rolling suitcase with small wheels, a tall lighthouse with beam lines, a rounded airplane, a palm tree with a face in the trunk, a camper van, a sun disk with short rays, a hot air balloon with a small basket, a backpack with visible straps, a triangular tent, a cruise ship with stacked decks, and a camera with a circular lens.Ensure even spacing, consistent visual weight, uniform eye placement, and a bold, clean, graphic poster-like composition.*" **Practical tip:**\ For logo and vector design, prompts should define: 1. Graphic type (logo, icon set, symbol system) 2. Shape logic (geometry, symmetry, silhouette clarity) 3. Color system (strict palette definition) 4. Line discipline (consistent stroke, no texture) 5. Layout structure (centered, grid-based, scalable) 6. Constraints (no gradients, no shadows, etc.) Avoid texture or material-focused language. Vector output responds to structural definition and geometric clarity. Image ## Graphic Design: Posters vol. 1 This is where Recraft V4 becomes especially powerful for designers.\ **It demonstrates a clear understanding of:** Typographic hierarchy\ Visual weight\ Poster-scale composition When layout logic is explicitly described, the model responds with accurate and predictable results. **Examples** Minimal Poster System Large-format poster. Large-format contemporary graphic design poster. Large-format contemporary magazine cover .
"*Large-format poster combining a cinematic desert landscape with bold graphic overlay. Background: open arid plain with warm sand tones, distant mountains, dramatic sky with soft evening light. Subtle print texture. Oversized electric-yellow abstract character drawn with thick expressive lines. A floating cloud-shaped figure with stretched limbs, one arm raised holding a symbolic lightning bolt. Simple facial features, closed eyes, calm expression. Flat solid fill, no gradients. Top left — minimal geometric logo mark + wordmark WILDFORM in clean modern sans-serif.Arched headline integrated with illustration: move slow move loud.Bottom line: Not perfect. Just alive. Strong contrast between realistic landscape and flat graphic layer. Contemporary, expressive, slightly rebellious poster aesthetic.*" "*Large-format contemporary graphic design poster combining raw editorial typography with expressive childlike crayon illustration. Warm off-white background with subtle paper grain and a strict grid layout defined by thin black frame lines and vertical divisions. At the top, an oversized tightly kerned bold black word in heavy grotesque sans-serif reads “BRAVE.” The center features a naive crayon drawing of a large sun and abstract landscape, made with thick wax strokes in saturated yellow, red, sky blue, and grass green, with uneven pressure and energetic scribbles. The drawing overlaps the grid and typography. Handwritten crayon text near it reads “dream big.”At the bottom, another oversized bold word reads “IMAGINE.” Between the illustration and lower headline, a narrow vertical column of small clean sans-serif text states: “Creative practice lives between structure and instinct. Design is not perfection — it is expression.” Footer in small uppercase sans-serif: “ARCHIVE / PRINT / POSTER / STUDIO / WORKS.”Brutalist black typography meets playful crayon art, flat colors, strong contrast, subtle paper texture — structured yet expressive and slightly rebellious.*" "*Large-format contemporary magazine cover combining bold naive painting with oversized modern typography. Background features a hand-painted still life of oversized seasonal vegetables — heirloom tomatoes, sliced purple cabbage, yellow squash, green zucchini, and red chili peppers — rendered in simplified flat shapes with thick matte acrylic texture and visible brushstrokes. The palette is saturated yet slightly muted, including dusty coral, mustard, sage green, deep plum, and soft cream tones. At the top, an oversized lowercase white wordmark in a heavy rounded sans-serif reads “harbor,” tightly spaced and spanning nearly the full width. Beneath it, a small clean sans-serif line reads: “design · culture · craft · studio · print · living.” A bold corner badge states “SPECIAL EDITION.” At the bottom, an organic cream-colored bubble contains the teaser text: “a handmade toolkit, slow projects, and a fold-out seasonal guide.”Clean editorial hierarchy, strong contrast between typography and painterly background, textured brush detail, modern independent print aesthetic — naive art combined with confident contemporary graphic design.*" **Prompt Structure:** 1. Format and scale (poster, cover, large-format) 2. Background / visual layer A 3. Graphic or image layer B 4. Typographic hierarchy (what is the largest?) 5. Text placement logic (top, center, edge, curved, cropped) 6. Contrast between layers 7. Overall compositional mechanics Recraft builds the layout instead of randomly placing elements. **Practical tip**: for hybrid posters (photo + graphic): Define each visual language separately, then describe how they interact. ## Graphic Design: Posters vol. 2 **Examples** Experimental Graphic Poster Playful glossy 3D poster. Promotional poster. Contemporary fashion campaign poster.
"*Playful glossy 3D poster featuring three iconic rubber ducks styled as high-fashion characters in a strong centered vertical stack. Each duck wears an exaggerated outfit: oversized futuristic sunglasses with a metallic puffer vest, a dramatic feathered collar with chunky jewelry, and a sporty streetwear hoodie with tiny sneakers. Smooth semi-gloss rubber texture, soft studio reflections, rounded toy-like proportions, subtle highlights, and confident minimal expressions. Background is flat saturated baby-blue. Layered 3D sticker elements float around the ducks — chunky stars, lightning bolts, smiley faces, safety pins, hearts, speech bubbles, and glossy tags in hot pink, yellow, cobalt, and white — creating diagonal movement and depth.Typography is bold and dynamic: inflated glossy hot-pink bubble letters reading “DUCK MODE” arch across the top, slightly overlapping the upper duck. A large diagonal shiny cobalt 3D headline “RECRAFT SPLASH” cuts across the lower half. A tilted oval sticker reads “LIMITED DROP.” Below the stack, small clean sans-serif text says “Designed for loud personalities.” In the corner, a glossy red capsule logo reads “QUACK CLUB — Tokyo / Milan.” High-fashion parody meets collectible toy culture, saturated, sculptural, poster-ready.*" "*Promotional poster for “System Lab” designer tableware collection from Formgrid Studio featuring sculptural ceramics and typography in a fresh grass-green tone — slightly acidic but not neon — set against a deep matte black background for sharp modern contrast. Centered symmetrical composition, photographed straight-on with a slightly high angle, soft diffused studio lighting from the top-front creating smooth highlights and gentle shadows.An extremely large bold condensed sans-serif headline “DESIGNER TABLEWARE” spans the full width at the top, edge-to-edge and slightly cropped, with smaller “VAJILLA DE DISEÑO” beneath. Oversized “SYSTEM LAB” overlays the center with smaller “LABORATORIO DE FORMAS” below. Vertical left text reads “EDITION 05 / EDICIÓN 05.” Bottom-right rounded rectangle contains “360° COMPOSITION / 360° COMPOSICIÓN,” with “FORMGRID STUDIO / ESTUDIO FORMGRID” nearby.The arrangement includes a sculptural low table with S-shaped legs and circular cutout shelf, an organic shallow platter with five petal-like lobes, a tall dual-handle glossy pitcher, and three matte bowls stacked in descending size. Minimalist industrial aesthetic, bold hierarchy, acid-leaning green on black, clean contemporary poster design.*" "*Contemporary fashion campaign poster blending realistic street photography with hyper-detailed 3D characters and bold experimental typography. A wide pedestrian crosswalk in a modern city under soft daylight — detailed asphalt, crisp zebra stripes, natural shadows, cinematic grading. A young avant-garde stylist walks confidently left to right, slightly off-center, captured in documentary realism with wind in the hair and a layered experimental outfit of saturated tones, technical fabrics, metallic accents, and structured tailoring.Plush-tech hybrid mascots move in the same direction along the stripes — one leading, one strong in the foreground, two behind at staggered depths — with ultra-real fur, glossy plastic and chrome details, and precise shadows aligned to the crosswalk. Seamless blend of real photography and CGI.Bold layered typography: “URBAN PARADE” oversized across the top, partially cropped; “NEW SEASON” and “COLLECTIVE DROP” as strong blocks on the right; date “04.12 — 05.08” in perspective along a stripe; curved tagline “Walk your story forward.” near the ground; compact “CITYFORM LAB” logo in the corner. Add large glossy sticker graphics — circular “EXCLUSIVE DROP,” rectangular “STREET EDITION,” diagonal “LIMITED RELEASE.” Strong asymmetry, unified motion, high-fashion meets collectible culture, dynamic editorial poster energy.*" **Prompt Structure:** 1. Core scene or subject 2. Additional 3D / graphic elements 3. Typographic dynamics (overlap, perspective, layering) 4. Color energy 5. Depth and spatial layering 6. Overall movement and tension ***Logic:*** Define layers first → then describe interaction. Image ## Illustration Strength Recraft V4 is highly capable in stylized and narrative illustration. **Examples** Digital anime-style. Сolorful illustration. Bold stylized illustration.
"*Digital anime-style chest-up portrait on a solid black background, centered. An androgynous young teen with cool pale skin and soft lavender blush, short choppy black hair with subtle teal highlights and small metallic clips. Large grey-blue eyes with bright highlights, thin straight brows, small muted rose lips, calm introspective expression.Wearing a modern sailor-inspired top in deep charcoal with a subtle geometric pattern, off-white collar trimmed in cobalt, and an oversized cobalt bow. Ten stylized koi fish in saturated reds, blues, gold, pink, white, and black swirl around the head and shoulders, overlapping the hair and partially crossing the face; one bold black-and-gold koi curves across the lower face. High contrast, flat lighting with soft shading, clean edges, dreamlike and contemporary mood.*" "*Сolorful illustration of a creative art studio filled with instruments and posters, but with a slightly more painterly, hand-drawn feel rather than clean vector graphics. A short-haired woman with visible arm tattoos and round glasses gently holds a large fluffy dog. She wears a loose white shirt and a dark hat. Around her are guitars, vinyl records, speakers, cables, sketchbooks, and framed portraits of abstract animal faces layered on the walls.Warm mustard, brown, terracotta, and burnt orange palette with soft tonal variation. Forms remain simplified and graphic, but edges are slightly irregular, as if drawn with ink and brush. Add a very subtle, delicate paper-like grain and gentle watercolor-style shading to avoid a flat vector look. Soft layered color transitions, light uneven brush marks, and faint depth in shadows. Cozy, artistic, intimate atmosphere — playful yet slightly textured, warm, and handcrafted rather than digitally perfect.*" "*Bold stylized illustration of an androgynous basketball player mid-dribble in extreme exaggerated perspective, with disproportionately enormous legs dominating the composition and oversized sneakers pushed dramatically toward the viewer. The entire figure is fully visible and fills the frame, but the legs occupy most of the visual space, making the torso appear smaller above them. Strong foreshortening and dynamic twisting posture.Color palette taken from the reference: saturated lemon-yellow ground, intense sky-blue background, vivid orange-red sun accent, bright acid neon green sneakers as the dominant highlight color, lavender-purple laces, soft beige midsole details, charcoal-grey outsole, crisp white socks, and bold black outlines. High-contrast flat graphic blocks.Add smooth gradients across the massive legs and shoes for volume, subtle airbrushed shading beneath the sneakers, and a delicate speckled grain overlay to avoid a clean vector look. Slightly irregular hand-drawn outlines remain visible. Energetic, surreal, contemporary poster aesthetic with bold acid green focus and exaggerated proportions.*" **Prompt Structure:** 1. Drawing style (anime, painterly, graphic, exaggerated) 2. Main character and pose 3. Line behavior (clean, irregular, bold) 4. Color logic 5. Surface treatment (flat, grain, watercolor shading) 6. Depth structure (gradients, airbrush, shadow softness) 7. Emotional tone \ ***Logic:*** You define drawing logic, not camera logic. Image ## 3D & Dimensional Visual Strength Recraft V4 handles dimensional form with strong material and lighting awareness. **Examples** Ultra-detailed stylized 3D render. Cinematic stylized 3D night. Digital 3D fashion lookbook render.
"*Ultra-detailed stylized 3D render in a high-end designer toy aesthetic from a top-down bird’s-eye view. A small sculptural character with rounded geometric anatomy and ultra-smooth semi-matte resin surfaces reclines diagonally on a large flowing chaise lounge with an integrated wicker detail, creating a strong graphic composition.The look is extravagant: lavender and powder-blue layered shoulders, dusty-coral torso, mint-green glossy knee-high boots, chunky pastel sneakers in mint, aqua and coral, plus oversized bangles, elongated earrings, a woven beige conical hat, tinted glasses, and a flowing ribbon accent. Around the chaise are a slate-blue side table, a peach ceramic vase with pastel branches, blush flowers, a folded parasol casting a radial shadow, and a woven rug.Soft peach, lavender, dusty coral, mint, wicker brown and blush tones dominate. Bright directional daylight creates long defined shadows from above, preserving a clean, sculptural, pastel-balanced editorial 3D still life.*" "*Cinematic stylized 3D night scene with strong depth and contrast, rendered in ultra-smooth plastic-like materials with refined matte surfaces and gentle subsurface scattering. A small stylized girl with an oversized head rides a skateboard through a dim urban alley. She has a playful, slightly mischievous expression, heavy-lidded minimal eyes, rounded proportions, two bouncing hair puffs, and a tiny snug beanie. She wears an oversized puffy jacket, loose wide-leg pants, and chunky sneakers in brighter electric blue and luminous cobalt tones, clearly separated from the darker surroundings.The background is deep and moody — navy-to-indigo gradient sky, shadowed brick walls, posters fading into darkness. The alley recedes with clear atmospheric perspective: sharper foreground, cooler midground, darker softly blurred background. Dry tiled pavement with subtle reflections, no rain or wet surfaces. Soft distant neon glow\.Lighting is cinematic: strong cool key light from above-right, soft side fill, and a clean rim light outlining her silhouette. Visible cast shadows from the skateboard and legs enhance depth. Low eye-level camera angle, shallow depth of field, modern animated short-film aesthetic with clear spatial layering and dimensional contrast.*" "Digital 3D fashion lookbook render featuring six anthropomorphic characters arranged in a clean 3 × 2 grid on a single solid light grey background and matching floor, figures slightly smaller in frame with generous negative space, soft frontal diffused lighting and gentle grounded shadows. All share identical minimalist faces with very small glossy bead-like black eyes, no pupils, no nose, and a thin minimal mouth line. Chibi proportions with oversized heads and compact bodies, matte plush materials with subtle metallic shine on layered chains. All styled as contemporary rappers with oversized silhouettes and bold streetwear attitude: a cream-white hare in an electric orange tee and sage varsity jacket with gold chains; a coral horned creature in a cropped white bomber over an apricot hoodie with beige cargos and a heavy chain; a charcoal raccoon in a moss oversized hoodie with layered chains and relaxed trousers; a lavender cat in an emerald oversized coat over a terracotta tee with wide charcoal pants; a black panther with subtle velvety fur wearing a cobalt hoodie and acid yellow crossbody bag with wide beige pants and platform sneakers; and a sand-colored fox in an ivory color-block jacket over a dusty-orange shirt with layered chains and white sneakers. Bright saturated palette, cohesive fashion-forward hip-hop collectible aesthetic." **Prompt Structure:** 1. Render type (designer toy, cinematic 3D, lookbook) 2. Form and proportion system 3. Material behavior (matte, gloss, plastic, fabric) 4. Spatial environment (floor, background, atmosphere) 5. Lighting direction and intensity 6. Camera angle and depth 7. Color system ***Logic:*** If you only describe the object, the result is interpretive.\ If you define the physical system, the result becomes controlled Image ## Working with Text  Recraft V4 handles text better than most generative systems, especially when you describe hierarchy. **Examples** Bold urban lifestyle. Bold avant-garde café menu. Contemporary premium food.
"*Bold urban lifestyle editorial featuring three avant-garde middle-aged men of distinctly different ethnic and visual backgrounds, with radically unique faces — varied bone structures, skin tones, wrinkles, scars, facial hair, asymmetry — real, unpolished, lived-in presence with no harmony or idealization. They sit close together at a small city street café table: one mid-bite holding food casually, one making a clear assertive hand gesture toward the camera, the third slightly back, calm and observant. All maintain direct or near-direct eye contact, confident, grounded, unapologetic.Shot at eye level in natural daylight with a documentary street-photography feel, tight framing on faces, hands, and table, honest skin texture and imperfections visible, softly blurred background. Controlled color palette with deep black base, clean white contrast, and one single bold accent color (cobalt blue, emerald green, or burnt orange), flat high-contrast tones only.Oversized experimental editorial typography overlaps the image without covering faces, using one or two short manifesto-style statements such as “NO RULES,” “CULTURE FIRST,” or “NOT FOR EVERYONE.” Multiple bold sticker-style graphic elements in the accent color visually connect the trio. Mood: raw contemporary street culture meets high-fashion editorial — confident, intelligent, manifesto-like, not made to please, not for everyone.*" "*Bold avant-garde café menu design in a luxury editorial style, photographed in an elevated, design-forward table setting. A large-format folded or partially unfolded menu card dominates the frame as the clear focal point, printed on thick premium matte paper with crisp edges and precise folds, placed on polished stone, lacquered wood, or tinted glass. The headline “VERA MENU” appears oversized and highly legible in an expressive high-fashion sans-serif with sculptural yet clean letterforms. The content is grammatically correct, clearly structured, and professionally written in proper English, with large evenly spaced typography and no filler text: sections such as ESPRESSO (Single-origin espresso, Double espresso), BREAKFAST (Sourdough toast with cultured butter, Soft scrambled eggs with herbs), and DESSERTS (Dark chocolate tart, Vanilla custard with seasonal fruit).Use a vivid high-impact palette — saturated coral, electric pink, deep cobalt blue, warm orange, and creamy off-white — with strong color blocking only, no gradients, and a disciplined typographic hierarchy. Surround the menu with minimal sculptural avant-garde cutlery featuring exaggerated forms and refined finishes. Soft directional editorial lighting creates subtle shadows while keeping the menu surface and typography crisp and fully readable. Overall impression: avant-garde luxury dining where graphic design meets contemporary art — bold, refined, collectible, no real brands, no clutter, focused on scale, clarity, and sculptural composition.*" "*Contemporary premium food packaging photographed in soft natural daylight, stacked rectangular matte cardboard boxes with sharp edges held slightly off-balance in human hands for a natural feel. Each box is a single refined color — deep forest green, warm cream, muted sky blue, soft clay beige — creating strong contrast. Bold oversized sans-serif brand name in dark ink is confidently cropped by the edges, with one short supporting line only, e.g., “VERA — Artisanal foods crafted slowly.” All typography is large, clear, and fully legible, no micro text or decorative clutter. A minimal abstract organic symbol (seed, leaf, or stone shape) appears flat and slightly irregular on the lid.One box is partially open, revealing two or three small clear glass jars with matte metal lids and simple bold labels containing gourmet spreads like pâté or eggplant caviar, rich natural textures visible. Matte paper texture with subtle grain, crisp printing, clean background, shallow depth of field. Modern, warm, premium editorial aesthetic — tactile, calm, design-forward, and quietly luxurious with intentional clarity in every element.*" **Prompt Structure:** 1. Format type (menu, editorial, packaging, poster) 2. Primary headline (size + placement) 3. Secondary hierarchy 4. Color blocking logic 5. Print surface and material 6. Spatial placement 7. Readability constraints 8. Place all required text in quotation marks ***Logic:*** Define the typographic system — not just the phrase. Quoted text ensures precise text rendering within the layout. Image ## Designer Photorealism Recraft’s realism feels curated rather than chaotic. **Examples** Ultra high-fashion. Cinematic fashion editorial. A cinematic editorial photograph.
"*Ultra high-fashion avant-garde close-up editorial portrait set in a cold cinematic studio. Deep monochrome cobalt-blue foggy background with a smooth gradient haze and soft atmospheric depth, minimal environment with no visible edges. The single character appears sculptural and iconic, with realistic natural skin texture — subtle pores, matte finish, sharp cheekbones, icy blue eyes, and an emotionless expression. Hair is platinum-blonde in a sharp geometric bob. They wear a dramatic couture piece featuring an oversized architectural ruffled collar made of translucent electric-blue tulle and layered organza, forming wave-like sculptural shapes around the neck. Makeup is minimal with muted burgundy lips and a soft powder finish.Strong long-exposure motion blur creates flowing trails and ghosted silhouettes behind the body and collar, while the face remains perfectly sharp. Clean, refined aesthetic with smooth gradients, no grain or painterly texture, high-end studio lighting. Cinematic, museum-like atmosphere with surreal quiet tension — modern couture captured mid-motion.*" "*Cinematic fashion editorial photograph captured outdoors during golden hour, focused on wind, motion, and atmosphere. The camera sits very low to the ground, almost inside tall wild grass, shooting upward to create an immersive, dynamic perspective. Softly blurred wheat-like plants dominate the foreground, forming organic diagonal lines that naturally frame the imageA young woman stands above the camera in a relaxed three-quarter pose, body slightly turned away while her head turns back toward the lens with direct eye contact, caught mid-movement rather than staged. She wears an oversized red hooded garment made of lightweight flowing nylon, voluminous and dramatic with visible drawstrings. Strong natural wind inflates and wraps the fabric around her body while her hair blows across her face, partially obscuring it.Warm golden-hour light casts a soft glow, with the sky fading from warm yellow near the horizon to gentle blue above. The setting is an open untamed field with no urban elements. Shallow depth of field keeps her face sharp while foreground grass and distant background remain blurred. Poetic, powerful, and alive — raw fashion captured in real wind, no studio lighting or artificial posing.*" "*A cinematic editorial photograph of a black man standing still on a busy crossing platform while crowds rush past her in heavy motion blur; he remains sharp and calm as people dissolve into streaks of movement; slow shutter effect, long exposure photography; muted urban color palette, soft industrial lighting, concrete textures, metro train in background; feeling of isolation, introspection, emotional contrast between chaos and stillness; contemporary fashion editorial mood, minimal styling, natural expression; shallow depth of field, cinematic framing, realism, high detail.*" **Practical tip:**\ \ Avoid stacking dramatic or evaluative adjectives.Precision and concrete description produce more reliable results than exaggeration.\ Prompt structure is flexible. The sequence of elements can be adjusted to change emphasis and influence model behavior. Elements placed earlier in the prompt receive higher priority and may shape the result more strongly.\ Common structural patterns include: * Scene / subject → style → details(prioritizes composition and content before aesthetic treatment) * Style → scene / subject → details(prioritizes visual language and overall aesthetic before composition) Both approaches are valid. Choose the structure based on which aspect should dominate the output. Image **How to Apply This in Real** **Branding:** Geometry → hierarchy → spacing → scalability → constraints **Fashion / Campaign:** Light → material → framing → simplified background **Vectors:** Silhouette → shape clarity → system consistency **Exploration:** 3–6 word prompts → generate variations **Posters:** Grid → margins → text size relationships **The Core Principle** **Recraft V4 adapts to your level of clarity** **|** Short prompts → model designs with you.\ Long prompts → model executes your architecture.\ \ **The model is strong in both modes |** Prompt engineering here is not about verbosity. \ It’s about defining structure, surface, space, hierarchy, and control.\ \ The more intentional your visual thinking, the more consistent and refined your results. # Graphic design Source: https://www.recraft.ai/docs/prompt-engineering-guide/visual-formats/graphic-design Graphic design brings together images, text, and layout to communicate information clearly. Effective prompts emphasize structure, hierarchy, and consistent use of typography and color so the model can produce organized and readable compositions. ## Examples
"*Minimal poster design for jazz concert, bold yellow headline typography on black background, sans-serif lettering, high contrast*" "*Avant-garde surrealist magazine cover, model's portrait fragmented into floating mirror shards reflecting different dimensions, magazine title constructed from 3D letters dripping down image, impossible architecture emerging from model's hair, eyes replaced by cosmic nebulas, gravity-defying objects suspended in space, psychedelic color shifts, dreamlike Escher-inspired composition, cutting-edge experimental editorial design, museum-worthy surreal photography*" "*Bold 3D coffee advertisement, geometric low-poly coffee cup made of sharp triangular facets suspended in darkness, enormous brutalist sans-serif letters "DARK ROAST" dominating frame - typography so massive it bleeds off all edges, severe angular composition, monochrome black and espresso brown with electric orange geometric accent, Cinema 4D architectural rendering, minimalist maximalism, 65% oversized cropped typography 35% abstract 3D cup sculpture, contemporary luxury brand aesthetic, striking visual hierarchy*" # Illustration Source: https://www.recraft.ai/docs/prompt-engineering-guide/visual-formats/illustration Illustration allows for expressive, stylized, or imaginative interpretation. The medium and emotional tone play a central role in shaping the final look. This style supports a wide range of techniques, from watercolor to ink to digital painting. Specify the medium and emotional tone to guide the model toward the character and atmosphere you want. ## Examples
"*Watercolor illustration of close up child and fox walking in the forest, soft diffused edges, flowing pigment washes, warm orange and green palette, visible paper texture, dreamy atmosphere, painterly style*" "*Black ink illustration of child and fox in forest, bold line work, crosshatching shadows, high contrast, pen and ink technique, detailed linework, white paper*" "*Digital painting of child and fox in forest, smooth brushstrokes, layered blending, soft lighting effects, contemporary illustration style, vibrant colors, polished finish*" # Logos and icons Source: https://www.recraft.ai/docs/prompt-engineering-guide/visual-formats/logos-and-icons Logos and icons distill identity into simple, recognizable shapes that work at any scale. They depend on minimal detail, clear geometry, and a restrained palette. Direct prompts toward form and clarity to guide the model toward effective symbolic designs. ## Examples
"*Minimalist pasta brand logo centered in composition, circular icon with brand name integrated as negative space cutout within the circle, letters carved from pasta nest spiral, text visible through transparent areas of symbol, typography and icon fused as one shape, soft golden beige outline on white, modern serif letterforms becoming part of circular geometry, unified design where you cannot separate text from image, balanced central placement with white space buffer, zen minimalist aesthetic where form and meaning unite*" "*Vintage badge logo for artisan pasta brand, circular emblem with wheat sheaf and pasta shapes in center, ornate decorative border, ribbon banner with "Est. 1892", classic Italian typography, warm gold and cream palette, traditional family recipe seal design*" "*Line art icon logo for pasta brand, simple outline of spaghetti wrapped around fork, consistent stroke width, monoline style, clean and minimal, works at small sizes, modern culinary app aesthetic*" # Photorealism Source: https://www.recraft.ai/docs/prompt-engineering-guide/visual-formats/photorealism Photorealism relies on accurate lighting, believable textures, and clear perspective. Emphasizing these elements helps the model produce images that resemble real photography. The goal is to make the image feel captured rather than created. ## Examples
"*Photorealistic medium portrait of woman in tall grass field at sunset, framing from shoulders up with face and flowing hair in focus, hands completely out of frame below, golden wheat and wildflowers surrounding her at shoulder level, warm golden hour backlighting creating rim light on hair, soft bokeh background with rolling fields, f/2.8 shallow depth, natural glowing skin, peaceful dreamy expression, simple natural style, warm amber and green tones, serene pastoral mood, lifestyle editorial photography, 50mm portrait lens, cinematic natural light*" "*High-end perfume product photography, elegant crystal bottle positioned on single geometric black obsidian block floating in void, dramatic spotlight from directly above creating sharp defined shadow beneath, rest in complete darkness, one precise beam of golden light hitting cap creating starburst reflection, ultra-minimalist composition with maximum negative space, luxury meets brutalist aesthetic, sharp focus throughout with f/11, museum-quality lighting setup, sophisticated contemporary commercial style, powerful visual restraint, Vogue-level editorial perfume shot*" "*Photorealistic wide shot of massive waterfall cascading into emerald pool deep in tropical rainforest, lush green vegetation covering every surface, morning mist rising from water creating ethereal atmosphere, sun rays penetrating through dense jungle canopy above, vibrant ferns and moss-covered rocks in foreground, crystal clear turquoise water, dramatic scale showing towering cliff walls, natural landscape photography, cinematic depth and layers, rich saturated greens and blues, shot on 16mm ultra*" # Vector art Source: https://www.recraft.ai/docs/prompt-engineering-guide/visual-formats/vector-art Vector art depends on precise shapes and clean geometry. Because it does not rely on texture or lighting for realism, prompts should highlight structure and simplified visual form. A focus on form, alignment, and clear edges guides the model toward the structured look typical of vector graphics. ## Examples
"*Flat vector illustration of mountain landscape, simple geometric shapes, two-color palette (blue, green), clean lines, minimal design*" "*Flat vector illustration of mountain landscape with layers, geometric peaks, stylized trees, gradient sky, bold outlines, modern graphic design style, balanced color palette*" "*Detailed flat vector illustration of mountain landscape, layered geometric mountains, stylized forest elements, gradient sunset sky, clean vector paths, contemporary poster design, rich color palette with complementary tones*" # Choosing a model Source: https://www.recraft.ai/docs/recraft-models/choosing-a-model **V4** is the best default choice for most tasks. It delivers the highest design taste, prompt accuracy, and output quality across raster and vector. Use it for everyday generation and fast iteration. **V4 Pro** shares the same creative capabilities as V4 but outputs at higher resolution. Choose it when you need print-ready assets, large-format visuals, or pixel-dense detail. **V3** is a strong option when you need style consistency, precise text positioning, or artistic level control. **V2** remains valuable for specialized creative effects and unique style options that are exclusive to the model. # Recraft V2 Source: https://www.recraft.ai/docs/recraft-models/recraft-V2 Recraft V2 is an image generation model released in March 2024 and the first model trained from scratch by Recraft. With 20 billion parameters, it was a breakthrough in human anatomical accuracy and the first to support brand consistency and brand color inputs. It also introduced vector image generation (SVG output), as well as minimalistic icon and illustration styles. V2 is available both in the web tool and via the API. In API use, it has a lower inference cost than V3. While V2 does not match V3 in prompt adherence or text accuracy, it remains preferred by many users for a set of specific styles and vector-based workflows. As of 2025, V2 is the only model that supports generating vector icons with brand consistency, allowing multiple icons to be produced with consistent line width, corner shapes, and other stylistic rules. ### **V2 curated styles** The curated library in V2 includes **Plastic 3D**, **Linocut**, **Grain**, and **Vector Art**, among others, and remain favorites of Recraft designers. * **Plastic 3D** produces consistently stylized 3D figures and icons in a chosen color palette, and is often used for app UI and presentation graphics. * **Linocut** is a vector style that can be monochrome or palette-based, and is frequently used for packaging. * **Grain** is valued for its textured look, common in blog and book illustrations. * **Vector Art** is popular for its distinctive outputs and is often used for logos, despite V2’s lack of accurate text generation. ### **V2 icon and pictogram styles** V2 includes several icon styles with different behavior. The **Recraft V2 Icon** style produces a different style for each generation when creating icons one-by-one. To achieve consistency, a new Vector Icon style must be created, or the Image Set tool can be used with **Recraft V2 Icon** selected. The **Recraft V2 Pictogram** style works similarly: each generation returns a new style unless the Image Set feature is used. This style also supports adjustable parameters such as cap type (see in-tool for available options). Other V2 icon styles produce consistent results by default. The most popular is **Vector Icon Outline**, which generates outline-style vector icons with a consistent look. ### **Limitations of V2** Some style names exist in both V2 and V3, but the visual results can differ significantly between them. V2 does not support accurate in-image text generation and may produce typographic errors. It uses the **Level of detail** parameter, which is not supported in V3, and does not support the **Artistic level** parameter available in V3. # Recraft V3 Source: https://www.recraft.ai/docs/recraft-models/recraft-V3 Recraft V3 is an image generation model released in October 2024. It ranked first for five consecutive months on the Artificial Analysis public benchmark on Hugging Face, maintaining a clear lead over models from Midjourney, OpenAI, and others. V3 introduced major advances in photorealism and text rendering. It was the first Recraft model to generate mid-size text accurately and, as of 2025, is the only model capable of placing text at specific positions in an image. It supports a rich library of curated styles, an infinite library of AI-generated styles, and—like V2—can maintain style consistency from a small set of example images without retraining. V3 also includes the **Artistic level** parameter for adjusting the artistic fidelity of an image. Lower values produce more standard images, such as direct portraits, while higher values create less conventional compositions with more peculiar angles or perspectives. Because of its flexibility and quality, V3 is recommended for most use cases. # Recraft V4 Source: https://www.recraft.ai/docs/recraft-models/recraft-V4 Recraft V4 is the latest generation of Recraft's proprietary model family, released in **February 2026**. With a focus on visual and design taste, V4 produces images with balanced composition, cohesive color, and refined detail that feel intentional rather than stock-like, across both raster and vector outputs. Recraft V4 text-to-image models are available in four versions: * **Recraft V4**: Standard raster model (1024x1024), \~10 seconds per image * **Recraft V4 Vector**: Standard vector model, \~15 seconds per image * **Recraft V4 Pro**: Higher-resolution raster model (2048x2048), \~30 seconds per image * **Recraft V4 Pro Vector**: Higher-resolution vector model, \~45 seconds per image **Pro** delivers higher resolution and finer detail, perfecting anatomy issues and improving realism across complex scenes. It produces more reliable results for detailed compositions, print-ready visuals, and other high-fidelity applications. All versions of Recraft V4 are available on all plan types, including the Free plan. ## Exploration Mode Exploration Mode is available in Recraft V4 and allows you to generate multiple visual directions from a single prompt. Instead of producing one result, it returns a set of variations, helping you compare options and select the most suitable outcome more efficiently. This mode is particularly useful during the early stages of the creative process, when you are exploring styles, compositions, or concepts and want a broader overview before refining a specific direction. **How to Use Exploration Mode in Recraft** 1. **Open Recraft Studio** and start a new generation task. 2. In the generation panel, **select “Exploration Mode”** instead of the default mode. 3. **Enter your prompt** as you normally would (describing the idea you want to explore). 4. Adjust any additional settings you need ( aspect ratio, etc.). 5. **Run the generation**, and Recraft will produce **multiple visual directions** from that one prompt. 6. Review the set of outputs and choose the direction you want to refine further. **Exploration Mode** is designed to help you quickly see a range of creative options without needing separate prompts for each variation. You can then pick one to continue working on or refine with additional prompts. ## Key features **Design-forward visual judgment**: Produces images with balanced composition, cohesive color relationships, and controlled detail, resulting in a more stylish, modern look that can be applied directly in professional creative workflows. Outputs tend to feel deliberate rather than generic or stock-like, making V4 well suited for workflows where aesthetic quality matters alongside prompt accuracy. **Exceptional prompt understanding**: Interprets detailed creative direction, accurately following visual relationships such as reflections, materials, and background interactions. **Readable, structured text generation**: Produces clear, legible text for infographics, menus, signage, and packaging designs. Handles short and mid-length phrases with high fidelity to layout and typography. **Rich output variety**: Generate everything from stylized illustrations to photoreal images and production-ready vectors. V4 maintains visual coherence and high detail across diverse creative directions. **Sharper detail and higher realism**: Improved textures, color balance, and lighting deliver clean geometry and consistent visual depth, avoiding the flat or synthetic look common in stock-style imagery. **High-quality vector generation**: The only model capable of generating editable, production-quality **SVG** graphics. Vector outputs preserve scalable geometry and discrete color regions suitable for professional design workflows. **Higher native resolution (Pro version)**: Ideal for professional production workflows that require large formats or pixel-dense detail. The Pro model outputs at 2048x2048 px for design-grade clarity across print, product, and marketing use cases. **Supported formats** Export results in **SVG**, **PNG**, **JPG**, **PDF**, **TIFF**, and **Lottie**. **Default availability** All V4 models are available in the web app and via the API. ## Limitations The following features are not yet supported in Recraft V4: * Style creation * Prompt-based editing * Image sets * Artistic level control ## Availability Recraft V4 is accessible to all users via the **model selector** in the Recraft app and API, and replaces V3 for most generation tasks. # How to create a color palette Source: https://www.recraft.ai/docs/recraft-studio/color-palettes/how-to-create-a-color-palette You can create a custom palette in one of three ways: 1. **Add hex codes manually**\ Enter specific color codes from your brand guide or visual spec to control the palette with precision. 2. **Use the eyedropper tool**\ Sample colors from any image on the canvas using the color picker. This is helpful when referencing past generations or imported brand materials. 3. **Extract colors from an image**\ Upload an image from your computer and Recraft will automatically extract the dominant colors. This is ideal for building palettes based on moodboards, campaign assets, or photography. Palettes are saved to your library and can be edited, renamed, or deleted at any time. A small color indicator on the palette icon shows whether a palette is currently active. # Recoloring existing images Source: https://www.recraft.ai/docs/recraft-studio/color-palettes/recoloring-existing-images ## Raster images For raster images, you can’t change specific colors directly. However, you can adjust overall tone and feel by selecting **Adjust colors** in the context panel located on the right of the canvas. This gives you access to sliders for: * Color spectrum * Saturation * Brightness * Contrast * Opacity Recoloring is especially helpful for light retouching or shifting the mood of a generated photo or painting. ## Vector images Vector images offer more detailed color control. After generation, an **Adjust colors** options appears in the prompt panel. You can: * Reduce color count to simplify the image and improve print-readiness. * Recolor using a saved palette by clicking on one from your library. There are two modes for adjusting colors, accessible via labeled tabs within the Adjust color panel: * **Swatches** (default): Adjust individual colors by clicking and replacing them. * **Spectrum**: Adjust groups of similar colors all at once. Ideal for complex illustrations with subtle tonal variations. Tip: Spectrum mode is more efficient for batch changes or when fine-tuning a unified look across multiple similar hues. # Using color palettes in generation Source: https://www.recraft.ai/docs/recraft-studio/color-palettes/using-color-palettes-in-generation When generating images, you can apply a color palette to guide the output: * On the Free plan, Recraft offers a rotating selection of preset palettes. These palettes are hand-picked and shuffled for variety and quick inspiration. * On a paid plan, you can create and save custom palettes, which can be applied to any image generation. Palettes can be used with any image type: photorealistic, illustrated, vector, or icon. You can also apply a background color separately to control the image’s overall tone. To apply a palette, click the **Color Palette icon** in the prompt panel. You can also explore randomized inspiration palettes from the “More” menu, useful for experimenting with color direction. Tip: The more complex your prompt or style, the harder it may be for the model to strictly adhere to limited palettes. Simpler prompts and styles tend to yield more accurate color matching. # How to export an image Source: https://www.recraft.ai/docs/recraft-studio/exporting-and-sharing/how-to-export-an-image 1. Select the image or images on the canvas. 2. Click the **Export** button in context panel. 3. Specify your image dimensions, format (PNG, JPG, TIFF, PDF SVG, or Lottie) and resolution. 4. Click the **Export** button. You can also right-click the image and choose **Export as...** from the context menu. Vector images can be exported as SVG and PDF for scalable use. # How to share an image Source: https://www.recraft.ai/docs/recraft-studio/exporting-and-sharing/how-to-share-an-image You can share an image in multiple ways: * right-clicking on the in canvas and selecting **Copy image link** * opening the image in the Community, Favorites, or History panels and clicking on the **link icon** to copy the URL. You can then share the URL. A Recraft account is required to view shared images. If the image is public, the link will be viewable by others. On paid plans, you can also set privacy settings before sharing. # Upscaling Source: https://www.recraft.ai/docs/recraft-studio/format-conversions-and-scaling/upscaling Recraft Studio offers two upscaling options, Crisp and Creative, depending on your needs for resolution and detail. * **Crisp upscale** increases resolution and sharpness without altering the structure or content of the image. Ideal for icons, vector-style illustrations, and assets where visual consistency is critical. * **Creative upscale** enhances resolution while subtly regenerating image content. It can improve anatomy, textures, and fine details, making it well suited for photos, portraits, or group scenes. Note: all creative upscale results are slightly different, and some iterations might render better than others. Note: Creative Upscale is a more resource-intensive process than Crisp Upscale. If you only need to increase resolution, use Crisp Upscale for faster results. If you want to enhance small details, improve anatomy (such as faces or hands), or add definition to background elements, Creative Upscale is the more suitable option. **How to upscale an image**: 1. Select an image on the canvas. 2. In the context panel, click either **Crisp Upscale** or **Creative Upscale**, depending on the image type and desired result. 3. Wait for the upscaled version to appear on the canvas. 4. When you're satisfied, export the upscaled image in your preferred format (e.g., PNG, JPG, PDF). Note: Upscaling doesn't support transparent images. The result of the upscaled image with a transparent background will have a white background instead of a transparent background. # Vectorizing Source: https://www.recraft.ai/docs/recraft-studio/format-conversions-and-scaling/vectorizing You can convert raster images into editable vector graphics directly within Recraft Stuio. Once vectorized, the image becomes scalable without quality loss and fully editable in terms of color. **How to vectorize an image**: 1. Place a raster image (e.g., PNG or JPG) onto the canvas. 2. Select the image, then click the **Vectorize** icon in the context panel. 3. Use the **Color Reduction** slider to limit the number of colors in the output—this helps simplify complex images or prepare for print. 4. Edit colors using two available modes: * **Swatches mode** (default): Click on individual color swatches to replace specific colors. * **Spectrum mode**: Adjust entire ranges of related colors at once—for example, shift all blue tones to green. Ideal for images with many subtle variations. 5. When you’re satisfied, export the vector as an SVG file for use in web, print, or design workflows. **Note**: Not all images can be cleanly vectorized. Photorealistic images, 3D renders, or highly textured illustrations may produce artifacts when converted to SVG, since the format supports only simplified shapes, lines, and fills. # Inpainting Source: https://www.recraft.ai/docs/recraft-studio/frames/inpainting Inpainting lets you regenerate specific parts of an image while keeping the rest intact. This is useful for correcting details, replacing elements, or making creative changes without starting from scratch. Inpainting is currently available in the old interface, which is accessible by clicking the **Old interface** toggle in the top navigation. **How to use inpainting**: 1. Select an image on the canvas. 2. Click the **Edit area** button in the top toolbar. 3. Choose a selection tool: **Lasso**, **Brush**, **Area**, or **Wand**. 4. Highlight the part of the image you want to change. 5. Enter a new prompt describing the desired update. 6. Click the **generation button** to apply the edit. The selected area will be replaced based on your prompt while the rest of the image stays the same. **Note**: When using inpainting on large images, Recraft automatically downscales the image before applying edits, which can result in a noticeable loss of quality in the final output. ## Tips for better results * If the selected area is too small, the change may be subtle or unnoticeable. Select a larger area for more effective inpainting. * Inpainting works best on images resolutions of 1024 pixels or smaller. For larger images, Recraft automatically downscales before editing. The result will be returned at the supported resolution of 1024 pixels. * After multiple rounds of inpainting, image quality may begin to degrade. If that happens, try reverting to an earlier version and batch several edits together in a single step. * To learn more about the selection tools available, visit [Selection tools](/recraft-studio/work-area/selection-tools). # Introduction to frames Source: https://www.recraft.ai/docs/recraft-studio/frames/introduction-to-frames Frames are flexible containers that allow you to compose, extend, or combine multiple image elements. They’re especially useful for designing layouts, expanding images to fit different aspect ratios, or creating content like social posts and promotional graphics. **How to use a frame**: 1. From the toolbar, click the **+ symbol** and select **Frame** to add a frame to your canvas. 2. Drag an existing image into the frame, or create a new one inside it. 3. Add additional elements, such as text or other visuals, directly into the frame if needed. 4. Select the frame and write a prompt describing the desired result. 5. Click **Generate frame** to generate a new composition that incorporates the image and elements inside the frame. Frames support outpainting (generating around an existing image); multi-element composition; and layout generation that follows the selected Recraft style whether it's photorealistic, illustrative, or vector-based. # Outpainting Source: https://www.recraft.ai/docs/recraft-studio/frames/outpainting Outpainting expands your image by generating new visual content beyond its original borders. It’s especially useful for changing aspect ratios, extending compositions, or adapting existing images to new formats. ### How to use outpainting (two methods): **Method 1**: 1. Double click on an image on the canvas. 2. A resizable frame will appear around the image. Click and drag the gray bars on any or all sides of the frame to your desired new size. 3. Click the **Expand** button. 4. Optionally, add a prompt in the box below to edit or enhance your image. **Method 2**: 1. Add a **Frame** to your canvas. 2. Drag an existing image into the frame. 3. Leave the prompt unchanged, or add a short prompt if you want to influence the expansion. 4. Click the **generate button** in the prompt panel to generate new content around the original image. 5. Optionally, add additional elements (like text or overlays) inside the frame to include them in the final composition. # History Source: https://www.recraft.ai/docs/recraft-studio/history All images you generate, including images from deleted projects, appear in both the **History page** and the **History panel**. You can access History in two ways: * From the **Recraft home menu** in the top navigation, select **History** to open the full [History page](https://www.recraft.ai/history). * From within a project, click **History** in the top navigation to open the History panel. The History page and the History panel expose the same underlying image history, but they serve different purposes. * The **History page** is a full management view for browsing, inspecting, and organizing images. * The **History panel** is a lightweight, in-context view for quickly reusing images while working on a project. ## History page vs History panel ### History page The **History page** is the primary place to manage your generated images. It provides full functionality, including: * filtering images by type * searching across your entire history * viewing detailed image metadata * copying prompts and image links * downloading images * deleting images * selecting multiple images for bulk actions Use the History page when you need to review, organize, or clean up your image history. ### History panel The **History panel** opens inside a project and is designed for quick access and reuse. It supports: * searching your image history * dragging images onto the canvas * inserting a copy of an image into the current project The History panel does not support filtering, bulk actions, or image deletion. For full management and inspection, open the History page. ## Browsing and searching (History page) ### Filters You can filter images on the History page by: * All images * Photorealism * Illustration * Vector art * Icon ### Search Use the search bar to find images using simple keywords, such as *red panda* or *rainbow cake*.\ Search matches against the prompt used to generate the image. ## Working with images (History page) ### Hover actions When you hover over an image thumbnail, you can: * Copy the prompt * Add the image to Favorites * Download the image * Delete the image * Select the image for bulk actions (a checkbox appears in the corner) ### Image detail panel Click any image to open its image detail panel. This panel shows: * the prompt used to generate the image * whether the image is raster or vector * the aspect ratio * the model or style used From the image detail panel, you can also: * add the image to Favorites * copy the image URL * download the image * permanently delete the image ## Bulk actions on images (History page) Bulk selection allows you to perform actions on multiple images at once. ### Selecting multiple images 1. Open the [History](https://www.recraft.ai/history) page. 2. Hover over a thumbnail to reveal the checkbox in the corner. 3. Click the checkbox to select the image. 4. Continue selecting additional images as needed. * A bulk action bar appears at the bottom showing how many images are selected. 5. Click a checkbox again to deselect an image. 6. Click the **X** in the bulk action bar to clear all selections. ### Bulk delete, download, or favorite Once one or more images are selected, the bulk action bar lets you: * **Download** all selected images * **Favorite** all selected images * **Delete** all selected images (confirmation required) **Note**: Bulk delete cannot be undone. You can still regenerate images later using their original prompts. ## Using images in your canvas From the **History panel** inside a project: * Drag an image onto the canvas to place it directly * Or click a thumbnail to insert a copy into your current project Images added from History behave like any other canvas image and can be edited, transformed, or combined with other elements. # Background tools Source: https://www.recraft.ai/docs/recraft-studio/image-editing/background-tools ## Remove background Recraft Studio’s Remove background tool is especially tuned for AI-generated images, making it more effective than the background removal features found in most other AI image generators. You can remove the background from any image with one click, creating a transparent version for compositing or mockups. To remove a background, select the image and click **Remove background** in the context panel. The background will be deleted, leaving only the subject. This is especially useful for preparing product overlays or isolating graphics. Remove background functionality works with images with resolution of 1024 by the smaller side. For images with large resolution there might be some artifacts on the border of the image. **Note**: When using Remove background on large images, Recraft automatically downscales the image before applying edits, which can result in a noticeable loss of quality in the final output. ## Change background **How to change the background of an image**: 1. Select a generated or uploaded image. 2. Select the image, then click **Change background** in the context panel. 3. Wait for the background area to be masked. 4. Enter a new prompt for the background in the prompt panel, then click the **generate button**. 5. Your image with new backgrond will appear on the canvas as a new image. # Editing images with natural language Source: https://www.recraft.ai/docs/recraft-studio/image-editing/editing-images-with-natural-language Recraft Studio supports targeted image editing using integrated external models **Google Nano Banana**, **OpenAI GPT‑4o**, **Seedream**, **Flux Kontext**, and **Qwen-Image** that respond to natural language prompts and optional reference images. These models are built directly into the Recraft interface, so you can make precise edits without switching platforms or starting from scratch. Natural language editing allows you to modify specific aspects of an image (like clothing, background, or facial expression) while preserving the overall structure and style. Only the elements described in your prompt are changed; everything else remains intact. ## Use cases * Make small adjustments, such as changing clothing, background elements, or facial features * Refine visual details without regenerating the entire image * Maintain stylistic or character consistency across multiple images * Add, remove, or modify specific objects within a scene * Adjust visual elements like lighting, color palette, or seasonal context * Update product features or character traits while keeping composition intact * Apply a consistent style or mood using reference images ## How to edit an image with natural language prompts 1. Open a project and upload or generate the image you want to modify. 2. Select the image on the canvas. 3. In the prompt panel, select a model: * **Nano Banana** (normal, Pro, Pro 4K) * **Seedream V4** * **GPT‑4o** (Low, Medium, or High quality) * **Flux Kontext** (Pro or Max quality) * **Qwen-Image** 4. Enter a prompt describing the change (e.g., *“make the jacket red”* or *“change background to nighttime”*). 5. (Optional) Click the **paper clip icon** within the prompt panel to attach reference images: * Up to **9 images** for GPT‑4o, Nano Banana, and Seedream * **1 image** for Flux Kontext and Qwen-Image Reference images allow the model to incorporate visual elements like style, character, pose, or layout into the output. In your prompt, describe what you want to borrow from each image and how you'd like them composed. 6. Click the **generate button** to apply the edit. 7. You can repeat the process as needed. Each new edit builds on the previous result without requiring a full regeneration. ## Supported models Currently, the reference image feature is supported only by select external models. You can manually choose the model, or GPT‑4o Medium will be used by default when a reference image is attached. Recraft standard image generation is included in the table below for reference. | Model | Credits | | ------------------------------- | ---------- | | Recraft raster image generation | 1 credits | | Recraft vector image generation | 2 credits | | Flux Kontext Pro | 2 credits | | Flux Kontext Max | 4 credits | | GPT-4o Low | 1 credit | | GPT-4o Medium | 6 credits | | GPT-4o High | 24 credits | | Nano Banana | 2 credits | | Nano Banana Pro | 15 credits | | Nano Banana Pro 4K | 30 credits | | Seedream V4 | 2 credits | | Qwen-Image | 2 credits | ## Notes on external models Each external model has its own supported aspect ratios. For example, GPT‑4o supports only 1:1, 3:2, and 2:3 formats. If your original image uses an unsupported aspect ratio (e.g., 16:9), the output will default to the closest supported format. The credit cost for using external models depends on each model's API pricing. # How to crop an image Source: https://www.recraft.ai/docs/recraft-studio/image-editing/how-to-crop-an-image To crop an image, hold ⌘ Cmd (Mac) or Ctrl (other systems) and drag its border to your desired dimensions. # Manual editing tools Source: https://www.recraft.ai/docs/recraft-studio/image-editing/manual-editing-tools **Manual selection tools** Recraft Studio provides several selection and modification tools for precise manual edits, such as the **Lasso**, **Brush**, **Area selector**, **Magic wand**, **Eraser**, and **Cut** tools. These allow fine control over what parts of an image are affected when applying edits, regenerations, or style changes. For detailed descriptions and usage of each tool, go to [Selection tools](/recraft-studio/work-area/selection-tools). ## Edit area You can manually select and edit an image or design using the Edit area feature. To activate it, select an image on the canvas, then click **Edit area** in the context panel that appears at the right of the canvas. This will open a panel where you can choose the Lasso, Brush, Area, and Wand tools. To exit these selection tools, click **Edit area** again. ## Duplicate an image There are multiple ways to duplicate an image: * Use the copy-paste function: Cmd/Ctrl+C and Cmd/Ctrl+V * Hold Option on Mac and Alt on other OS, then click and drag the new image * Right-click the image and select **Duplicate** # Raster effects Source: https://www.recraft.ai/docs/recraft-studio/image-editing/raster-effects Recraft Studio effects let you add visual treatments to any **raster image** on the canvas, such as illustrations, photos, or painted textures. These effects do not apply to vector images. You can combine multiple effects, reorder them, or remove them at any time. Effects are non-destructive; you can remove them without modifying the original image. ## Accessing raster effects Raster effects are accessed through the **Adjust colors** control in the context panel. To open raster effects: 1. Select a **raster image** on the canvas. 2. In the right sidebar (context panel), click **Adjust colors**. 3. Click **+** in the **Effects** section to add an effect, such as **Blur**, **Drop Shadow**, or **Stroke**. The Effects section is available only when a raster image is selected. ## Drop shadow Drop shadow adds depth by placing a soft shadow behind your image. To add a drop shadow: 1. Select a raster image on the canvas. 2. In the right sidebar, open the **Effects** section and choose **Drop Shadow**. 3. Adjust the available controls: * **Position (X, Y):** Offsets the shadow horizontally and vertically * **Blur:** Controls how soft or sharp the shadow edge appears * **Color:** Sets the shadow color (use darker tones for natural shadows) Tip: Drop shadow also works with images with transparent backgrounds. ## Stroke Stroke adds an outline around the frame of your raster image. To add a stroke: 1. Select a raster image. 2. In **Effects**, choose **Stroke**. 3. Customize your outline using: * **Weight:** Controls the thickness of the outline * **Color:** Sets the outline color * **Position:** * **Inside** – stroke sits just inside the image boundary * **Outside** – stroke sits outside the image boundary * **Center** – stroke is placed half inside, half outside the boundary ## Blur Blur softens the entire image for atmospheric or stylistic effects. To apply a blur: 1. Select a raster image. 2. In **Effects**, choose **Blur**. 3. Adjust the **Strength** slider to increase or decrease the blur intensity. Blur is frequently used to create depth, soften textures, or reduce detail behind foreground elements. ## Removing an effect Each effect added appears as an entry under **Effects** in the right sidebar. Click **–** next to any effect to remove it. Removing an effect immediately resets all of its settings to default. # Remix Source: https://www.recraft.ai/docs/recraft-studio/image-editing/remix Remix generates new versions of an image while keeping the overall concept intact. Each remix introduces slight differences in elements like character appearance, object position, or scene composition. The Remix function does not use a prompt. Instead, it uses the visual content of the original image to generate alternatives in different aspect ratios. **How to use Remix**: 1. Select an existing image on the canvas. 2. Click on the **Generate variations** in the context panel. 3. Choose the number of images and aspect ratio for the new image(s). This is optional, but ideal for creating variations of an image in the same aspect ratio. 4. Click **Generate** to produce a set of new outputs based on the selected image. # Agentic mode Source: https://www.recraft.ai/docs/recraft-studio/image-generation/agentic-mode Recraft Studio's Agentic mode enables users to talk and generate images in Recraft Studio like a design assistant. To activate Agentic mode, select **Agentic** in the generation controls in the prompt panel. ## What's possible with Agentic mode **Chat-driven generation & iteration**: Start by describing an idea or intent, and iterate on the visuals using the entire chat context. **Brainstorming prompts and design ideas.** Use the Agentic mode interface to help craft and refine prompts or explore different design directions without complex instructions. **Step-by-step workflows**: Multi-step tasks are guided automatically to help move from brainstorming to finished assets in fewer steps. **Smart tips in context**: Get suggestions for next actions, such as upscaling images to higher clarity and resolution, discovering different styles, or improving results within a workflow. **Canvas integration**: Result appears side-by-side on the canvas, preserving context to help compare or build on earlier outputs without resetting. ## How credits work in Agentic mode Agentic mode uses the same credit system as the rest of Recraft Studio. Each interaction with the language model consumes credits, in addition to any credits used for image generation or editing. * Simple, one-off requests may use as little as 1 credit. * Longer conversations with multiple iterations and accumulated context consume more credits. * Agentic credits are separate from image generation and editing costs. For better control over usage, start a new chat when switching tasks instead of continuing a long conversation. Credit usage scales with compute: * Quick requests stay lightweight. * Complex, multi-step workflows use more credits. This keeps everyday tasks efficient while accurately reflecting the cost of more advanced sessions. ## Example Agentic mode workflows * Start by asking for help with a prompt for a design task, e.g.: *“I need a hero illustration for a SaaS landing page about teamwork. Help me craft 3 prompt ideas”* * Select one of the suggested prompts, edit it as needed and try generating in different styles by typing *“generate it in flat style”* or *“generate it in photorealistic style”.* * To change specific details on an image, type in chat e.g., *“take 2nd image generated in flat style, but add a dog showing pet-friendly office”* * Experiment with styles by attaching a reference image, then ask in chat to apply the style to another image * When generating or iterating on images, use chat to specify aspect ratio, number of images, or to apply Recraft features such remove/replace background, vectorize, create variations, or upscale. * Try using some more advanced design workflows or chain of generations and editing actions e.g.,: * *“Generate a photoreal product shot of a ceramic mug on a plain light background. Then remove the background. Then produce a clean vector interpretation of the silhouette.”* * Use key visuals as image reference(s), then ask to create a marketing assets (such as a landing hero, social cover, email hero), specifying desirable formats, colors or other important details. # Artistic levels Source: https://www.recraft.ai/docs/recraft-studio/image-generation/artistic-levels Artistic levels allow you control the complexity, composition, and visual tone of generated images. It is available in the Recraft V3 model. Settings range from **Simple** to **Eccentric**, offering six levels of control: * Simple levels tend to be clean, minimal, and straightforward, similar to stock-style visuals. * Eccentric levels become more dynamic and expressive, with varied layouts, richer compositions, and creative flourishes. * The levels in between offer increasing degrees of visual complexity, allowing you to calibrate the output to suit your project’s tone and context. **How to use the artistic level slider** 1. Start a new image by entering a prompt. 2. Locate **Artistic level** by clicking on the settings icon in the prompt panel. 3. Click the drop down menu and select your desired setting: Simple, Clear, Artistic, Expressive, Rich, Eccentric. 4. Click the **generate button** to generate your image using the selected visual tone. Tip: Set the Artistic level to **minimum** if you want people to look directly at the camera. # Aspect ratios Source: https://www.recraft.ai/docs/recraft-studio/image-generation/aspect-ratios Recraft Studio supports several preset aspect ratios that can be applied when generating new images. Use the **Aspect ratio selector** located in the prompt panel to choose the desired format. Aspect ratio affects how content is framed and composed, and is especially important when preparing assets for platforms with fixed dimensions, such as social media, thumbnails, or print layouts. **How to generate an image using a custom or unsupported aspect ratio** 1. Generate an image using the aspect ratio closest to your desired one. 2. Create an empty Frame and manually type in the **W** and **H** values to the exact dimensions you need. 3. Drag and drop the generated image into the Frame. 4. Click **Generate frame** to outpaint the empty space and fill the frame to your target size. ### Using the Aspect slider to crop an image You can crop an existing image by selecting it and using the Aspect slider to crop it to your desired dimensions. # Context panel Source: https://www.recraft.ai/docs/recraft-studio/image-generation/context-panel The context panel surfaces actions and properties for the current selection on the [canvas](/recraft-studio/work-area/canvas). It changes dynamically based on what is selected and provides a focused place to inspect, modify, and act on one or more objects. The context panel is selection-driven. It appears only when at least one object is selected on the canvas and updates automatically as the selection changes. ## What the context panel is responsible for The context panel handles object-specific and batch actions, including: * displaying properties for the current selection, such as dimensions and metadata * exposing actions that apply to the selected object or objects * enabling batch operations when multiple objects are selected * providing access to export and arrangement controls It does not create new content or interpret natural language instructions. Those actions are handled by the [prompt panel](/recraft-studio/image-generation/prompt-panel). ## Context panel states The context panel adapts based on selection state and object type. Available actions vary depending on whether a single object or multiple objects are selected, and on the types of objects in the selection. **Single object selected** When a single image or object is selected, the context panel shows properties and actions specific to that object. This may include: * object dimensions (width and height in pixels) * prompt and generation-related metadata (when applicable) * object-specific actions such as export, background removal, vectorization, or upscaling This state is used for inspecting and acting on individual elements. **Multiple objects selected** When multiple objects are selected, the context panel switches to batch actions. The set of available actions depends on the types of objects selected. For standard images, batch actions may include: * running the same prompts across all selected images * background removal * vectorization * crisp or creative upscaling * merging layers * exporting selected images For other object types, such as mockups, fewer batch actions may be available. In these cases, the context panel typically includes: * merge layers * arrange * export In all multi-selection cases, an **Arrange** control is available. This control allows you to group and rearrange selected objects along a horizontal or vertical plane. ### Actions for a single selected object When a single image or object is selected, the context panel may include the following actions: * **Reuse in a new image**: Use the selected image’s prompt and generation settings as a starting point for creating a new image. * **Generate Remix**: Create multiple variations of the selected image while preserving its overall structure. View [Remix](/recraft-studio/image-editing/remix). * **Remove background**: Automatically remove the background from the selected image. View [Remove background](/recraft-studio/image-editing/background-tools#remove-background). * **Change background**: Replace the background using natural language instructions. View [Change background](/recraft-studio/image-editing/background-tools#change-background). * **Edit area**: Apply natural language editing to a specific region of the image. View [Edit area](/recraft-studio/image-editing/manual-editing-tools#edit-area). * **Extract prompt**: Retrieve the prompt and settings used to generate the image for reuse or refinement. View [Extract prompt](/recraft-studio/image-generation/working-with-text-and-prompts/extract-prompt). * **Use in mockup**: Place the selected image into a mockup for product or scene visualization. View [Mockups](/recraft-studio/mockups/how-to-create-a-mockup). * **Vectorize**: Convert the selected raster image into an editable vector format. View [Vectorizing](/recraft-studio/image-editing/format-conversions-and-scaling/vectorizing). * **Crisp upscale**: Increase image resolution while preserving structure and detail. View [Upscaling](/recraft-studio/image-editing/format-conversions-and-scaling/upscaling). * **Creative upscale**: Increase image resolution while enhancing details and textures. View [Upscaling](/recraft-studio/image-editing/format-conversions-and-scaling/upscaling). * **Adjust colors**: Modify the color palette of the selected image. View [Color palettes](/recraft-studio/color-palettes/using-color-palettes-in-generation). * **Export**: Export the selected object in supported formats. View [How to export an image](/recraft-studio/exporting-and-sharing/how-to-export-an-image). ### Actions for multiple selected objects When multiple objects are selected, the context panel provides batch actions, including: * **Align**: Align selected objects to an edge (left, right, top, bottom, center). * **Arrange**: Adjust the relative ordering of selected objects as a group. * **Merge objects**: Combine selected objects into a single image. * **Export**: Export all selected objects together. View [How to export an image](/recraft-studio/exporting-and-sharing/how-to-export-an-image). # External models Source: https://www.recraft.ai/docs/recraft-studio/image-generation/external-models External models let you utilize top AI models directly in Recraft Studio, so you can generate and edit images without switching tools. | Model | Creator | Description | Time | Credits | | ---------------------- | ----------------- | --------------------------------------------------------------------------------------------------------- | ---- | ------- | | Flux 2 Max | Black Forest Labs | Maximum consistency for complex, multi-image workflows | 20s | 18 | | Flux 2 Dev | Black Forest Labs | Open-weight model for fast experimentation and pipelines | 6s | 3 | | Flux 2 Pro | Black Forest Labs | Production-grade realism with strong prompt alignment | 10s | 10 | | Flux 2 Flex | Black Forest Labs | Configurable quality, speed, and typography control | 20s | 13 | | Flux 1 Dev | Black Forest Labs | Developer-focused version for experimentation | 9s | 7 | | Flux 1.1 Pro | Black Forest Labs | Refined Flux model with high-fidelity rendering | 6s | 8 | | Flux 1 Kontext Pro\* | Black Forest Labs | Context-aware Flux variant preserving semantic alignment in scenes | 7s | 8 | | Flux 1 Kontext Max\* | Black Forest Labs | Advanced Kontext version with enhanced resolution and cinematic detail | 12s | 15 | | GTP Image 1.5 Low\* | OpenAI | Fast, lightweight layout-aware image generation | 15s | 2 | | GTP Image 1.5 Medium\* | OpenAI | Balanced quality and consistency for general use | 25s | 5 | | GTP Image 1.5 High\* | OpenAI | High-detail images with precise layout control | 50s | 24 | | GPT Image 1 Low\* | OpenAI | Lightweight GPT-4o variant for text-to-image tasks, multi image editing | 15s | 2 | | GPT Image 1 Medium\* | OpenAI | Balanced GPT-4o variant with improved detail and realism, multi image editing | 35s | 8 | | GPT Image High\* | OpenAI | High-quality GPT-4o setting delivering photorealistic outputs, multi image editing | 60s | 30 | | HiDream I1 Dev | HiDream AI | Experimental diffusion model optimized for artistic compositions | 12s | 6 | | Ideogram V3 Turbo | Ideogram AI | Fast inference mode optimized for speed | 9s | 6 | | Ideogram V3 Default | Ideogram AI | Standard setting balancing speed and quality | 12s | 11 | | Ideogram V3 Quality | Ideogram AI | Highest-fidelity Ideogram variant | 12s | 17 | | Imagen 4 | Google DeepMind | Advanced Imagen with higher realism | 8s | 8 | | Imagen 4 Ultra | Google DeepMind | Top-tier Imagen for ultra-photorealistic rendering | 10s | 13 | | Nano Banana\* | Google Gemini | Standard-resolution outputs, reliable generation, and foundational editing. | 12s | 7 | | Nano Banana Pro\* | Google Gemini | Faster generation, higher fidelity, accurate multilingual text, and advanced creative controls | 15s | 32 | | Nano Banana Pro 4K\* | Google Gemini | High-resolution variant delivering sharper detail and 4K-grade image outputs. | 30s | 63 | | Seedream V4.5\* | ByteDance | Structured, reference-driven high-fidelity image generation | 30s | 8 | | Seedream V4\* | ByteDance | Next-gen diffusion model delivering vivid colors, fine detail, and artistic realism | 20s | 7 | | Qwen-Image\* | Alibaba | Multimodal model built for high-fidelity image generation and text–image alignment | 7s | 4 | | Z-Image Turbo | Alibaba | Ultra-fast photorealism with strong bilingual (English/Chinese) text rendering and solid prompt adherence | 3 | 1 | \*Used for [editing images with natural language](/recraft-studio/image-editing/editing-images-with-natural-language) within Recraft. # Generating multiple images Source: https://www.recraft.ai/docs/recraft-studio/image-generation/generating-multiple-images ## Images per generation You can generate between 1 and 4 images per prompt by wihthin the image number selector in the prompt panel. This is useful for exploring multiple variations at once without needing to re-enter your prompt. The selected number of images will appear side by side on the canvas after generation. ## Image sets Image sets enable you to generate six visuals at once, each from a different prompt, while keeping them stylistically consistent. This feature is particularly useful when working with icons, illustrations, or batch assets for UI, marketing, or product design. **How to create an image set**: 1. From the **Canvas**, click the **+ symbol** then choose **Image set** from the toolbar. 2. Enter up to **six individual prompts**, each describing one image (e.g., "Sun icon", "Moon icon", etc.). 3. Select a style from the curated library or your saved/custom styles. You can also adjust the image settings (avoid text and negative prompt) colors. 4. Set the artistic level and aspect ratio as needed. 5. Click **Generate** in the Image set prompt panel to generate the full set. Once the set is generated, you can refine individual results by clicking on any image in the set and updating its prompt or parameters. This regenerates only the selected image while leaving the rest of the set unchanged. # Generation modes Source: https://www.recraft.ai/docs/recraft-studio/image-generation/generation-modes Recraft Studio offers two ways to generate images: **Agentic mode** and **Manual mode**. Each mode is designed for a different workflow, and you can switch between them at any time. Agentic and Manual modes are available in the new interface. If the **Old interface** toggle is enabled, these modes are not accessible. ## Agentic mode [Agentic mode](/recraft-studio/image-generation/agentic-mode) uses an LLM-powered assistant to help generate and refine images through a conversational workflow. In this mode: * The system understands context and intent across multiple turns, and keeps track of your history. * You can describe ideas naturally, refine them, and ask for suggestions. * The assistant may suggest adjustments, apply styles, or select models automatically when helpful. * You don’t need to restate details repeatedly as you iterate. * Image generation and edits follow the flow of the conversation. Agentic mode is best for exploration, iteration, and situations where you want guidance or are still shaping what you want to create. It may also be faster in *achieving your goal*, especially when you’re not sure how to prompt something. ## Manual mode Manual mode is a manual, prompt-based workflow with no conversational context. In this mode: * Each request is treated independently. * No history, context, or memory is applied between generations. * You control all inputs explicitly through the prompt and settings. * Generation is faster and more predictable. Manual mode is ideal when you know exactly what you want, need precise control, or are making quick, one-off generations. ## Switching between modes You can switch modes directly in the prompt panel: 1. Locate the mode selector in the prompt panel. 2. Select **Agentic** for conversational, context-aware generation. 3. Select **Manual** for manual, stateless prompt-based generation. Switching modes does not reset your project or canvas: * Existing images remain unchanged. * You can explore ideas in Agentic mode, then switch to Manual mode to fine-tune details. * You can move between modes freely as your workflow changes. ### Return to the old interface (optional) If you re-enable the **Old interface** toggle, the new interface is disabled and Agentic mode is no longer available. ## When to use each mode | Mode | Best for | Behavior | | ----------- | -------------------------------- | ------------------------------------------------------------- | | **Agentic** | Exploration, iteration, guidance | Context-aware, conversational, may suggest or adjust settings | | **Manual** | Precision, speed, manual control | Stateless, prompt-only, fully explicit | Both modes work within the same canvas and support Recraft Studio’s full set of image generation and editing tools. # How to generate a logo Source: https://www.recraft.ai/docs/recraft-studio/image-generation/how-to-generate-a-logo 1. **Create a new image**. Open your project and add a new image frame on the canvas. 2. **Select a vector style**. For the best results, choose one of Recraft’s vector models and styles designed for logo creation, such as Geometric Logo, Sharp Drawn Logo, or Playful Typographic. You can find more by typing "logo" in the style search bar. 3. **Write your prompt**. Describe the logo you want to create, for example: “logo ancient Greek sculpture, clever negative space”. To include text in the logo, place it in quotation marks. For instance, typing “Broccoli” will add that word as stylized text. 4. **Click the generate button**. Your logo will be generated in seconds. # How to generate an image Source: https://www.recraft.ai/docs/recraft-studio/image-generation/how-to-generate-an-image 1. Create a new project, or go to [http://recraft.new](http://recraft.new). 2. Describe your image in the prompt panel at the bottom of the canvas, then click the **generate button**. After a few seconds, the new image will appear on the canvas in a bounding box. ### Image generation options Use the buttons in the prompt panel to adjust how your image is generated: * **Agentic** / **Manual**: Select between [Agentic and Manual modes](generation-modes). * **Model**: Choose which image model to use. Different models are optimized for different output types, such as raster images or vector graphics. * **Style**: Browse curated styles or explore an infinite feed of AI-generated styles. You can apply a predefined style or create your own using reference images. * **Ratio**: Set the aspect ratio of your image to control its dimensions and framing, such as square, landscape, or portrait formats. * **Count**: Choose how many images to generate from a single prompt, up to 4. Multiple images are generated as variations and appear together on the canvas. * **Color**: Apply a predefined or custom color palette to guide image generation. # Prompt panel Source: https://www.recraft.ai/docs/recraft-studio/image-generation/prompt-panel The prompt panel is where image generation and natural language editing begin in Recraft Studio. It provides access to generation modes, models, styles, and settings, and adapts based on whether you are creating new content or editing an existing selection. ## What the prompt panel controls The prompt panel handles all prompt-driven actions, including: * generating new images, frames, mockups, and image sets * editing selected images using natural language * selecting models, styles, and generation modes * configuring generation parameters such as aspect ratio and number of images It does not manage layout, selection, or object-specific actions such as alignment, arrangement, or export. Those are handled on the [canvas](/recraft-studio/work-area/canvas) and in the [context panel](/recraft-studio/image-generation/context-panel). ## Prompt input and controls The prompt panel includes controls for configuring how images are generated or edited: * **Prompt field**: Enter natural language descriptions or instructions * **Generation mode**: Choose how prompts are interpreted (for example, Agentic or Manual) * **Model selection**: Select the model used for generation or editing * **Style selection**: Apply visual styles to generation * **Aspect ratio**: Control image dimensions * **Images per generation**: Generate multiple outputs at once Some controls are available only in specific modes or when editing an existing selection. ## Prompt panel states The prompt panel changes behavior based on selection state. **No selection** When nothing is selected, the prompt panel is in generation mode. You can: * enter a prompt to generate new images * choose a model and style * set aspect ratio and number of images * start new generations This is the default state for creating new content. **Single object selected** When a single image or object is selected, the prompt panel switches to editing mode. You can: * modify the prompt associated with the selected image * apply natural language edits or refinements * regenerate variations based on the current selection This state is used for iterating on existing images rather than creating new ones. **Multiple objects selected** When multiple images are selected, the prompt panel switches to **reference mode**. In this state: * the selected images are treated as **reference images** for a new generation * any prompt you enter is interpreted in relation to the selected images * the next generation creates a new image informed by those references, rather than editing them directly This state is used to generate new content guided by visual examples on the canvas. ## Docking and placement The prompt panel can be toggled between different positions depending on your workflow: * docked at the bottom of the canvas * docked to the side of the workspace using the prompt panel toggle Docking affects layout only. All prompt panel functionality remains the same regardless of position. # Selecting a style Source: https://www.recraft.ai/docs/recraft-studio/image-generation/selecting-a-style By default, images you generate are created by top-ranked Recraft’s V3 model. If you want to generate an image in a different style, click the **Style** button in the prompt panel to open the Style library. You will see two primary style libraries: * The **Curated** tab contains styles categorized by style type: Photorealism, Illustration, Vector art, Vector icon, Graphic design, Vector logo, and Other styles. * The **AI Generated** tab contains an "infinite" collection of AI-generated styles Select and set your style, then click **Generate button** to see the results. For more about styles, go to [Styles overview](/recraft-studio/styles/overview). # Adding text to an image Source: https://www.recraft.ai/docs/recraft-studio/image-generation/working-with-text-and-prompts/adding-text-to-an-image Recraft Studio lets you add and manipulate text directly on the canvas, either as part of your layout or as input for generation. Text can be layered manually over any image, used within a **Frame** to guide generation of images with text, or incorporated into prompts for image generation. ## Typing text into the prompt In Recraft Studio, you can generate text as part of an image by simply including it in the prompt. This is useful when you want the text to appear stylized within a design, such as a logo, sign, label, or poster, rather than adding it manually afterward. You can write prompts that include both a visual description and the desired text, or prompts that only describe the text itself. **Examples:** * `"A coffee shop logo with the words 'Java Joint'"` * `"The phrase 'Open super late' in neon sign style"` * `"Tote bag design with the words 'Support your local bookworm'"` Recraft Studio’s models will interpret the text prompt and incorporate it directly into the image, styled to match the overall look. You can increase accuracy by placing the text in quotation marks, which helps the model recognize it as literal text rather than a text description. This method is best suited for cases where the text should feel embedded in the visual design rather than overlaid. ## Adding text manually Manual text allows font selection from Google Fonts, color, size, line height, and letter spacing adjustments. This is best used for static overlays (e.g., captions, annotations, headlines) where you want full manual control. How to manually add text: 1. On the canvas, select the **text tool** from the left toolbar. 2. A blank cursor appears. Click anywhere on the canvas and start typing. 3. Use the **text context panel** to adjust typeface, font size, weight, line height, letter spacing, color, and alignment as needed. 4. Use the bounding box controls to resize, rotate, or reposition the text as needed. 5. You can combine text with other visuals, including images, mockups, or frames. ## Using text in a frame for AI layout generation For more advanced compositional control: 1. Add your image to the canvas. 2. Create a **Frame** and place the image and any text elements inside it. 3. Select the frame, then write a prompt (e.g., "Design a layout using the image and title text"). 4. Click **Recraft Frame**. Recraft Studio will use both the visual and text elements as input, generating a new composition with the text integrated according to the chosen style. This is useful for designing social media graphics, posters, or promotional layouts where you want AI assistance with composition and visual harmony. **Notes and limitations** * You can include multiple text elements within a Frame to simulate layout or multi-line designs. * Recraft Studio uses the position of text boxes in a Frame as layout suggestions. During generation, text may be repositioned or restyled based on the selected style. * Text added via a Frame becomes part of the generated image and cannot be edited afterward. * For vector images, any text included will be preserved in the exported SVG file. * Recraft V3 supports generating images with text, but does not handle very small text reliably. If text appears distorted or contains typos, try increasing its size. * Recraft V2 does not support image generation with text. Using this model may result in text rendering errors or misspellings. * Recraft V3 supports text generation within the following supported characters: '!', '"', '#', '\$', '%', '&', '\\'', '(', ')', '\*', '+', ',', '-', '.', '/','0', '1', '2', '3', '4', '5', '6', '7', '8', '9',':', '\<', '>', '?', '@','A', 'B', 'C', 'D', 'E', 'F', 'G', 'H', 'I', 'J', 'K', 'L', 'M', 'N', 'O', 'P', 'Q', 'R','S', 'T', 'U', 'V', 'W', 'X', 'Y', 'Z','\_', '', 'Ø', 'Đ', 'Ħ', 'Ł', 'Ŋ', 'Ŧ', 'Α', 'Β', 'Ε', 'Ζ', 'Η', 'Ι', 'Κ', 'Μ', 'Ν', 'Ο', 'Ρ', 'Τ','Υ', 'Χ', 'І', 'А', 'В', 'Е', 'К', 'М', 'Н', 'О', 'Р', 'С', 'Т', 'У', 'Х','ß', 'ẞ'.\ Notes: * Characters in different cases (like “B” and “b”) are treated as the same symbol. * Characters that appear identical are different characters with minute variations. * Characters not listed in this character set will not render. # Extract prompt Source: https://www.recraft.ai/docs/recraft-studio/image-generation/working-with-text-and-prompts/extract-prompt **Extract prompt** lets you generate a descriptive text prompt from any image, whether it was made in Recraft Studio or uploaded from elsewhere. The tool analyzes the visual content and returns a detailed description of what it “sees,” including aspects like subject, composition, technique, and style. This is especially useful when you want to: * Understand how to describe an existing image in prompt form * Recreate or iterate on a similar image * Explore stylistic attributes for reference or learning * Reuse successful results in new contexts * Refine previous generations without starting from scratch * Analyze how a particular outcome or style was achieved Unlike reusing a generation’s original prompt, **Extract prompt** works as an AI interpretation. It does not recover the exact text that was originally used to generate the image. #### How to use Extract prompt 1. Upload an image from your computer or select one on your canvas. 2. Click **Extract prompt** in the context panel. 3. Recraft will analyze the image and display a text description summarizing its content and visual style. 4. You can edit this description or use it directly as a new prompt for generation. *Tip:* Extract prompt works best with clear, well-defined visuals (illustrations, product photos, icons, etc.). The richer the image detail, the more specific and useful the extracted prompt will be. # Negative prompts Source: https://www.recraft.ai/docs/recraft-studio/image-generation/working-with-text-and-prompts/negative-prompts Negative prompts help guide the AI by specifying what you don’t want to appear in the image. This gives you more control over the outcome, especially when avoiding unwanted objects, styles, or features. **How to use negative prompts**: 1. Select your image on the canvas and click the **settings button** in the prompt panel. 2. In the **Negative prompt** field, enter the elements or concepts you want to exclude from the image. 3. Optionally, you can disable text generation in your output by toggling the **Avoid text in image** switch to **Yes**. 4. Generate your image as usual. Recraft will attempt to avoid the elements you specified. Tip: avoid double negation in negative prompts. For example, to exclude apples from an image, write `"apples"` in the negative prompt, not `"no apples"`. Using “no” can confuse the model and lead to the opposite effect. # Using custom fonts Source: https://www.recraft.ai/docs/recraft-studio/image-generation/working-with-text-and-prompts/using-custom-fonts You can upload your own font files and use them inside Recraft when adding text. Custom fonts are particularly useful for brand typography, logo refinement, packaging mockups, marketing visuals, and consistent presentation design. ## How to upload a custom font 1. Select the **Text** tool. 2. Go to the Context panel on the right, locate **Typeface** and click the **upload icon**. 3. Select your font file (`.woff`, `.woff2`, `.ttf`, or `.otf`). 4. The font will appear in your font list. 5. Apply it to any text layer. ## Important notes * Uploaded fonts are available within your workspace. * If a font file does not include certain character sets (for example, Cyrillic characters), those characters will not render correctly. * Ensure the font file contains all required glyphs for your language and use case. # About importing Source: https://www.recraft.ai/docs/recraft-studio/importing/about-importing Recraft Studio lets you bring your own images into the canvas. Imported images can be used across the platform’s tools, from mockups and vectorization to prompt-based editing and fine-tuning. Once imported, your image can be used to: * Generate a variation of the original visual * Use as a reference for prompt-based generation * Remove the background * Vectorize raster images (e.g., PNG to SVG) * Remix or edit specific areas with prompts * Apply your image to a product mockup * Extend the image using outpainting (frame tool) * Create a custom style for consistent generation * Upscale to increase resolution and detail **Notes on imported and edited images** * Only **original generated images** are saved to project history. Images that have been edited e.g., background removed or fine-tuned) will not appear in the project timeline. * Imported images, whether added to the canvas or uploaded to create a custom style, are not used for training AI models. * To reuse edited images, first **export them**, then **re-import** as needed. * When importing an image, Recraft Studio may default to an external model to enable natural language editing. # How to import an image Source: https://www.recraft.ai/docs/recraft-studio/importing/how-to-import-an-image There are a few ways to import an image: * Drag and drop an image directly from your system file browser onto the canvas * Click the **Upload image** button in the **toolbar** * Click **History** in the top navigation to open the history panel. Click on any image to place it onto the canvas. Once placed, an imported image will behave like any other canvas element. You can resize, rotate, and reposition it as needed. **Note**: Imported images are not used for training AI models. # Size limits and supported formats of imported images Source: https://www.recraft.ai/docs/recraft-studio/importing/size-limits-and-supported-formats JPG, PNG, WebP: * Width and height dimensions should not exceed 5800 pixels each * Max resolution: 16 megapixels * Max file size: 48MB SVG: * Max file size: 48MB # Interface language Source: https://www.recraft.ai/docs/recraft-studio/interface-language Recraft is designed for creators around the world. To make the experience more comfortable and accessible, the interface is now available in **six languages**. You can easily change the interface language directly from your **Profile settings**. ### How to change the language 1. Open your **Profile** in Recraft. 2. Locate the **Language** dropdown menu. 3. Select your preferred language from the list. Once selected, the interface will automatically update. ### Available languages Currently, Recraft supports the following languages: * **English** Image * **Deutsch (German)** Image * **Español (Spanish)** Image * **Français (French)** Image * **Українська (Ukrainian)** Image * **Русский (Russian)** Image You can switch languages anytime from the **Profile → Language** dropdown menu, and the change will apply instantly across the interface. # Interface overview Source: https://www.recraft.ai/docs/recraft-studio/interface-overview This video walkthrough highlights the two available chat layouts, explains how Manual and Agentic modes differ, and shows where to configure models, styles, and generation settings. It also demonstrates how generated images connect to on-canvas tools such as background removal, edit area, vectorization, and mockups.