Canva releases its first image generation model
TechCrunch Ivan Mehta
Canva launched its first image model, Canva Craft. It’s built to make cleaner text, sharper layouts, and easier edits than most image bots.
Based on reporting by TechCrunch, Ivan Mehta — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Canva has released its first image generation model, Canva Craft, in a direct bid to compete with the image systems from Google, OpenAI, ByteDance, and Meta. The company says the model was built by its research team and is now available inside Canva’s platform through the Dream Lab tab. It will also show up through the Canva API, with support planned for third-party chatbots such as ChatGPT and Gemini via Canva’s MCP.
The pitch is not just “make pictures.” Canva says Craft works in two stages. First, it reads a prompt and builds a plan that includes spatial relationships, relative scale, visual hierarchy, and text details. Then it generates an image at up to 2K resolution in as many as 30 aspect ratios. The company says that setup should help the model handle object detail and legible text better, while also making precise edits possible without disturbing the rest of the image.
There’s a twist in how it was trained. Instead of leaning on photographs, Canva says it trained Craft on design, using human-made templates from its own library. That includes posters, social posts, presentations, infographics, menus, and other text-heavy layouts. Because of that, Canva says the output can be turned into layered designs through Canva Magic, its image-to-design transformer. The company already released a model last year that understands designs and layers.
Canva also says the new system should help it save on API costs, while giving it more control over quality and speed. That matters because the company has relied on a mix of image models inside its product. Canva bought Leonardo.ai in 2024, and says researchers from that team helped build Craft. Two cheaper versions are on the way as well: Craftfast and Craft-ultrafast.
The timing is interesting. Canva was on track for $4 billion in revenue last year, but it cut its 2026 growth forecast, and external analysts reportedly pushed its valuation down from $42 billion to $35 billion. A house model does not fix everything, but it does give Canva more control over a core feature instead of renting the future from someone else.
My take — AI-written commentary, not fact-checked reporting
Canva is doing the obvious thing: stop depending on everyone else’s models and build your own. That’s the adult version of AI strategy, even if it’s less glamorous than posting benchmark confetti. The real test now is whether design-native generation beats generic image spam, because the internet has plenty of that already.
Read more about this at: TechCrunch
