Canva's first in-house AI image model, built from scratch by Canva Research and tuned for design output — legible in-image text and layout-aware graphics rather than photorealism.
Last verified · 2026-10-22 · by Moe Ameen
Canva Craft is Canva's first proprietary image-generation model, announced on October 22, 2026, and — per Canva — the first image foundation model built and trained from scratch by Canva Research. The notable thing isn't just that Canva now has its own model instead of renting one; it's what the model is tuned for. Most image generators are trained predominantly on photographs and are best at photorealism. Craft was trained on millions of human-made design templates from Canva's library — posters, social posts, presentations, infographics, menus, and typography-rich layouts — so its edge is design: readable, correctly placed in-image text and deliberate composition, the exact things photo-trained models tend to mangle.
Architecturally, Craft runs in two stages. It first interprets the prompt and drafts a plan — mapping spatial relationships, relative scale, visual hierarchy, and the text that will appear — before generating any pixels, which Canva describes as "less about pixels and more about planning." It outputs at up to 2K resolution across a wide range of aspect ratios, and results are editable and controllable rather than one-shot. Canva reports that Craft debuted fourth on the independent DesignArena leaderboard, behind three OpenAI models, making Canva (by its own account) the second-highest-ranked image lab behind OpenAI — treat that as early, company-reported framing.
Access and pricing: Craft is available now to Canva users via the Dream Lab tab — coverage at launch described it as open to all Canva users globally, though Canva's existing free/paid tiers gate usage by credits, so expect heavier use to favor paid plans — and to developers through Canva's public API; a Canva MCP Server that lets AI agents call the model is described as coming soon. Canva has also said it will soon ship two cheaper variants, Craft-fast and Craft-ultrafast, pitched as quality generation "at a fraction of the cost of leading frontier models." No per-image pricing was published at launch, so confirm current terms on canva.com.
The honest framing: Craft is an image model, not a content operation. It generates a still — a very good, text-legible, on-layout still — and stops there. It does not generate or clip video, write captions in your brand voice across formats, assemble a finished multi-slide post or blog, or schedule and publish anywhere. That's the line to keep in mind when you plan the workflow around it.
Craft solves the picture; it doesn't solve the publishing. Its whole advantage — legible text and deliberate layout — makes it the best kind of still to hand to a content engine, because a graphic that already reads well only needs to be made on-brand, turned into finished posts, and shipped. That's exactly where [Kompozy](/) picks up. Generate a poster, thumbnail, or background in Craft, then bring it into Kompozy, where your [Persona Brief](/glossary/persona-brief) governs the caption voice and banned words and [HyperFrames](/glossary/hyperframes) renders brand-exact Carousel Posts, Photo Posts, and Quote Graphics around the image — one look, one voice, every platform. Kompozy then schedules and publishes the set across the eight social platforms plus blog and email on [Autopilot](/glossary/autopilot), behind a per-post review gate.
The bigger reason to pair them is motion and format coverage. Craft makes stills; a feed needs more than stills. Because Kompozy is a full generation engine, the same campaign that opens with a Craft graphic can carry into face-locked [Persona Shorts](/glossary/persona-shorts) presented by a recurring avatar, vertical [Clipped Shorts](/glossary/clipped-short) cut from a long video, a Blog Article, and an Email Newsletter — all from one source, in one voice. Think of Craft as the still-image specialist in your stack and Kompozy as the engine that turns any asset into a scheduled, multi-format, on-brand week. If you also want a freeform design canvas to hand-tune a hero asset, the broader [Canva comparison](/alternatives/canva) covers that side.
Canva Craft is Canva's first proprietary AI image model, announced October 22, 2026, and built from scratch by Canva Research. It is tuned for design — legible in-image text and layout-aware graphics like posters, social posts, and infographics — rather than photorealism, and outputs up to 2K across a wide range of aspect ratios.
Text legibility is its headline strength. Canva says Craft was trained on millions of human-made design templates and plans the layout and in-image text before generating, which it credits for stronger, more readable text than photo-trained models typically produce. That makes it well suited to thumbnails, posters, and carousel slides.
Craft is available now via the Dream Lab tab in the Canva app — reported at launch as open to all Canva users globally, with Canva's usual free/paid credit tiers likely applying to heavier use — and through Canva's public API; it was not sold as a standalone per-image product. Canva has said cheaper Craft-fast and Craft-ultrafast variants are coming "at a fraction of the cost of leading frontier models," but hadn't published final pricing — confirm on canva.com.
No. Craft generates still images only. It does not generate or clip video, write captions across formats, assemble finished multi-slide posts or blogs, or publish to any platform. A content engine like Kompozy takes a Craft image and turns it into on-brand carousels, photo posts, and quote graphics scheduled across the eight social platforms plus blog and email.