كانفا تطلق Craft، أول نموذج لتوليد الصور تبنيه بنفسها، ومتاح لكل المستخدمينCanva launches Craft, its first in-house image model, available to all users

Canva has launched Canva Craft, the first image generation model built and trained from scratch by its in-house lab, Canva Research, and designed specifically for design. It is available from today to all Canva users worldwide through Dream Lab inside Canva, and to developers through Canva's API.

What's new? Instead of drawing the image straight from the prompt, it works in two stages: first it writes a plan for the composition, covering which elements appear, where, how big and what text must show, and then it renders the image at up to 2K resolution in more than 30 aspect ratios. Canva says this makes text clearer and placement more precise, and makes it easier to edit one part without changing everything.

The model was trained on millions of human-made templates from Canva's library, such as posters, social posts, presentations and menus. The company says it debuted fourth among image generation models on DesignArena's rankings, and first in the People & Portraits and Landscapes & Nature categories. You can write prompts in any language, and the model can create text in images in 55 languages.

But note: the DesignArena ranking is as reported by Canva in its announcement, and its internal evaluations were run by its own designers and researchers. The announcement did not say whether Arabic is among the 55 languages, and the company says it is still working on fine print and technical drawings.