


Qwen-Image is designed for sharp text rendering, multilingual support, and semantic image editing. Expect layouts that include paragraphs or fine-grained text (English and Chinese), accurate visual edits that preserve objects and context, and deeper visual understanding (object detection, depth, edges) built in.
It is an open-source 20B parameter MMDiT foundation model, specializing in precise text rendering (especially complex or multilingual text) and strong semantic editing performance.
Yes. The model supports TI2I (text + image→image edits) preserving visual fidelity and semantic meaning. Image editing retains context while allowing changes via text.
PNG, JPG, TIFF, PDF, and SVG for vectors from the Export panel.

Recraft gives you built-in access to leading external models directly from the canvas. Switch between Recraft’s own models and third-party options to refine images, edit with text prompts, and explore new creative styles.

Turn prompts into high-quality images and editable vector graphics with Recraft's state-of-the-art AI generator.

Enlarge visuals to increase resolution, add crisp detail, and remove artifacts. Make images print-ready with one click.