Image to image

Image to image creates a new image from an existing one. It can keep aspects of the original, such as composition, color, or subject identity, and change the rest based on your prompt. Use strength to set how similar the result stays to the original.

$0.04 / $0.08

per image with V3 raster / vector. V4 and V4.1 prices

Image + prompt → image

up to 10 MB, 16 MP

POST /v1/images/imageToImage

endpoint

Try it

Get an API token on your profile page (it needs a positive API units balance) and set it once per terminal session:

export RECRAFT_API_TOKEN=your_token

Then run. The request turns a public sample illustration into a winter scene. It uses the default model, recraftv4_1:

curl https://external.api.recraft.ai/v1/images/imageToImage \
  -H "Authorization: Bearer $RECRAFT_API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "image_url": "https://cdn.prod.website-files.com/655727fe69827d9a402de12c/679b82f54ef9617939df9d8e_motorcycle--going-fast.png",
    "prompt": "the same scene in winter, snow on the road",
    "strength": 0.4
  }'

# pip install requests
import os

import requests

response = requests.post(
    'https://external.api.recraft.ai/v1/images/imageToImage',
    headers={'Authorization': f"Bearer {os.environ['RECRAFT_API_TOKEN']}"},
    json={
        'image_url': 'https://cdn.prod.website-files.com/655727fe69827d9a402de12c/679b82f54ef9617939df9d8e_motorcycle--going-fast.png',
        'prompt': 'the same scene in winter, snow on the road',
        'strength': 0.4,
    },
)
if not response.ok:
    raise SystemExit(f'{response.status_code}: {response.text}')
print(response.json()['data'][0]['url'])

// Node.js 18+, run as an ES module (file.mjs)
const response = await fetch('https://external.api.recraft.ai/v1/images/imageToImage', {
  method: 'POST',
  headers: {
    Authorization: `Bearer ${process.env.RECRAFT_API_TOKEN}`,
    'Content-Type': 'application/json',
  },
  body: JSON.stringify({
    image_url: 'https://cdn.prod.website-files.com/655727fe69827d9a402de12c/679b82f54ef9617939df9d8e_motorcycle--going-fast.png',
    prompt: 'the same scene in winter, snow on the road',
    strength: 0.4,
  }),
});
if (!response.ok) throw new Error(await response.text());
const result = await response.json();
console.log(result.data[0].url);

# pip install openai
import os

from openai import OpenAI

client = OpenAI(
    base_url='https://external.api.recraft.ai/v1',
    api_key=os.environ['RECRAFT_API_TOKEN'],
)

response = client.post(
    path='/images/imageToImage',
    cast_to=object,
    body={
        'image_url': 'https://cdn.prod.website-files.com/655727fe69827d9a402de12c/679b82f54ef9617939df9d8e_motorcycle--going-fast.png',
        'prompt': 'the same scene in winter, snow on the road',
        'strength': 0.4,
    },
)
print(response['data'][0]['url'])

Response

{
  "created": 1759140000,
  "credits": 40,
  "data": [
    {
      "image_id": "0e6a1b8c-5b1f-4a8e-9f3a-2c7d4e8b9a10",
      "url": "https://..."
    }
  ]
}

data has one entry per image (n). Download each image from its url.
The model decides whether the output is raster or vector. A vector model returns SVG files.
Image URLs are temporary: files are kept for about 24 hours, so download them. See Limits and sizes.
credits is the number of API units charged for the request.
Pass "response_format": "b64_json" to get the image bytes inline, or "multipart" for the lowest latency. See Image inputs and results.

Parameters

Parameter	Type, default	Description
`image_url`	string, required in JSON	URL or data URL of the input image.
`image`	file, required in multipart	The input image, uploaded instead of `image_url`.
`prompt`	string, required	A text description of the areas to change. The maximum length depends on the model. See Prompt length.
`strength`	float, required	How much the result differs from the original, from `0` to `1`. `0` is almost identical, `1` is the least similar.
`model`	string, `recraftv4_1`	A Recraft V3, V4, or V4.1 model, raster or vector. V4.1 Flash is not supported. See All models for every `model` value.
`n`	integer, `1`	Number of images, 1 to 6.
`random_seed`	integer	Seed for reproducible results.
`controls`	object	Custom parameters to tweak generation. See Controls.
`style_match`	string	How closely the result matches the referenced style. Only `regular` is supported. A value the model doesn’t support is rejected. See Style match.
`response_format`	string, `url`	`url`, `b64_json`, or `multipart`. See Image inputs and results.
`image_format`	string, `webp`	`webp` (lossless) or `png`. Applies to raster output only. See Image format.
`style`	string	V3 only. A curated style, for example `Illustration`.
`style_id`	UUID	V3 only. Use a style as a visual reference. See Styles.
`negative_prompt`	string	V3 only. What the image must not contain.
`text_layout`	array of objects	V3 only. Words and their positions in the image. See Text layout.

The input image must be PNG, JPG, WEBP, or SVG, under 10 MB, up to 16 MP, with sides between 256 and 4096 px. See Input images.

Examples

Upload a local file

curl https://external.api.recraft.ai/v1/images/imageToImage \
  -H "Authorization: Bearer $RECRAFT_API_TOKEN" \
  -F "image=@image.png" \
  -F "prompt=the same scene in winter, snow on the road" \
  -F "strength=0.4"

import os

import requests

with open('image.png', 'rb') as image:
    response = requests.post(
        'https://external.api.recraft.ai/v1/images/imageToImage',
        headers={'Authorization': f"Bearer {os.environ['RECRAFT_API_TOKEN']}"},
        data={
            'prompt': 'the same scene in winter, snow on the road',
            'strength': 0.4,
        },
        files={'image': image},
    )
if not response.ok:
    raise SystemExit(f'{response.status_code}: {response.text}')
print(response.json()['data'][0]['url'])

Stay close to the original

A low strength keeps the result almost identical to the input, so the prompt changes only small details. Here n returns two variants.

curl https://external.api.recraft.ai/v1/images/imageToImage \
  -H "Authorization: Bearer $RECRAFT_API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "image_url": "https://cdn.prod.website-files.com/655727fe69827d9a402de12c/679b82f54ef9617939df9d8e_motorcycle--going-fast.png",
    "prompt": "golden evening light",
    "strength": 0.15,
    "n": 2
  }'

V3 with a style and a negative prompt

style and negative_prompt work only with V3 models. With recraftv3 an image costs $0.04. Use recraftv3_vector for SVG output.

curl https://external.api.recraft.ai/v1/images/imageToImage \
  -H "Authorization: Bearer $RECRAFT_API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "image_url": "https://cdn.prod.website-files.com/655727fe69827d9a402de12c/679b82f54ef9617939df9d8e_motorcycle--going-fast.png",
    "prompt": "the same scene in winter",
    "strength": 0.4,
    "model": "recraftv3",
    "style": "Illustration",
    "negative_prompt": "people"
  }'

Remix image

Make variations of an image without a prompt.

Enhance prompt

Expand a short prompt before you use it.

Recraft V4.1

The default model for this endpoint.

Pricing

Prices of every model and tool.

Get started

Models

Tools

Reference

$0.04 / $0.08

Image + prompt → image

POST /v1/images/imageToImage

Try it

Response

Parameters

Examples

Remix image

Enhance prompt

Recraft V4.1

Pricing

$0.04 / $0.08

Image + prompt → image

POST /v1/images/imageToImage

​Try it

​Response

​Parameters

​Examples

​Related

Remix image

Enhance prompt

Recraft V4.1

Pricing

Try it

Response

Parameters

Examples

Related