Canva’s Craft image model plans the layout before it paints

A layout plan comes first

Craft separates composition from rendering: it plans the arrangement before making the picture. Canva says the plan covers where objects and words go, how large they are and what should catch the eye first. A second stage produces the image.

Canva’s Magic Layers can then turn the image into editable elements. That is a separate conversion step; more advanced editing and the ability to work from reference images are still planned.

Canva’s original demonstration below shows the design ambition. The practical question is what happens when someone asks for a different headline or moves the product after the first draft.

A promising start in the image rankings

DesignArena’s image leaderboard placed Craft fourth when we checked on the morning of 23 October, behind three OpenAI models. People compare generated images and choose the result they prefer. That gives an outside view of how the pictures look.

It leaves a different question open: how easy are those pictures to turn into finished work? The ranking does not test brand rules, editable layers or a complete revision. A useful next test would follow the same real brief from first image to approved design.

What developers can use today

Teams building their own software can also use Craft through Canva’s image generation API, the connection that lets an app request an image. The current developer reference lists it as available, but the API remains in preview: it can change without warning, and public apps using it cannot yet pass Canva’s review for release to all users.

Canva staff advise developers to allow around a minute for a Craft image. Jobs use the user’s AI credits and return an image file. The main integration details are below for teams evaluating that route.