Canva Craft brings design-first image generation

Canva is moving beyond bundling third-party models toward a vertically integrated design AI stack, with editability and layout control as its clearest differentiators.

• –A two-stage planning-and-rendering architecture targets stronger composition, object placement, and readable text than generic image generators.
• –Craft outputs are designed to become editable layers, making the model more useful for real production work than a one-shot image tool.
• –Developers can call Craft through Canva’s API using `craft_v1_2k`, opening integrations for marketing, commerce, publishing, and design automation.
• –Canva says the model was trained on design-heavy data and evaluated by professional designers, a sensible strategy for posters, presentations, social graphics, and branded assets.
• –The real test will be whether Craft’s quality, latency, AI-credit pricing, and upcoming faster variants can compete with established image models.