Canva’s first in-house image model is now open to free accounts

Canva has released Craft, the first image-generation model its own research team built, and the company says it is available to all Canva users through the Dream Lab tab. Free accounts are included, but they draw on the same small monthly allowance of AI uses that Canva gives its Free plan: up to 20 a month. The release came on Oct. 22, 2026, and puts Canva’s research team in direct competition with the image models from Google, OpenAI, ByteDance and Meta.

AI, explained every day: The Deep View is the go-to daily newsletter for 600,000+ founders, knowledge workers and professionals who want to stay up to date with artificial intelligence.

Get The Deep View →

We may earn a commission if you sign up.

What Canva has not settled is how much of a free account’s monthly allowance a single Craft image uses. That unanswered detail decides whether the access is a trial or a working tool for a free user.

What Canva says Craft is

Canva’s own announcement of Craft says the model is “available now to all Canva users worldwide” through Dream Lab, the image-generation panel inside the editor, and that developers can reach it through Canva’s public API. The company describes a two-stage design. A planning model first drafts the composition from the prompt, covering which elements appear, where they sit, how large they are and what text they carry, and a rendering model then produces the image. Canva says this split is what makes targeted edits possible without disturbing the rest of the picture.

The announcement lists output at up to 2K resolution across more than 30 aspect ratios, with custom ratios allowed, and text rendering in 55 languages. Canva says it trained Craft on millions of human-made templates from its own library, such as posters, social posts, presentations, infographics and menus, rather than mainly on photographs, and refined it with feedback from professional designers who compared more than 25,000 image pairs against leading models.

TechCrunch’s report on the release adds that Canva acquired Leonardo.ai in 2024 and that researchers from that team contributed to the model. It also says two cheaper versions, Craft-fast and Craft-ultrafast, are in development, and that Canva plans to bring image generation to third-party chatbots through its Model Context Protocol server, which the newsroom page marks as coming soon.

The claim about free accounts

The statement that Craft is open to free accounts rests on two layers of sourcing. Canva’s announcement says “all Canva users” without naming plans, and TechCrunch’s report carries the same “all users” language without saying whether free accounts are included. The plan-level detail comes from Creative AI News, which reports that a Canva FAQ describes Craft as available to all Canva users, including Canva Free, with each image drawing on the account holder’s monthly AI allowance.

Canva’s own help page on understanding AI usage fills in what that allowance is. Canva Free gets “up to 20 uses of Standard AI tools, OR Premium AI tools” each month, with no access to Ultra tools. Standard tools use the least of that allowance, and once the limit is reached Standard and Premium tools slow down with short pauses between generations while Ultra tools stay locked until the reset. The allowance resets at midnight UTC on the first of the month. The help page does not mention Craft by name, so the connection between the allowance and the new model comes from the FAQ that Creative AI News reports rather than from the help page itself.

Creative AI News also notes a tier puzzle. Canva’s help article lists image creation as a Premium tool, while its pricing page lists it as Standard, and because Free shares one allowance across both tiers, the 20-use ceiling applies either way. Canva has not said which tier Craft belongs to, and this article does not assign one.

Canva’s ranking claim and the business case

Canva’s lead claim is about standing. The company says Craft debuted as the fourth most advanced image model on DesignArena, a benchmarking platform for AI-generated design, and that this makes Canva Research the world’s second most advanced AI lab for image generation. Its announcement adds category results: first in people and portraits among 87 models, first in landscapes and nature among 85, second in abstract and patterns, and fourth in products. These are Canva’s own reported rankings of its own model, and no independent test is cited. The announcement also says the model is covered by Canva Shield, the company’s set of input and output moderation, privacy and safety controls and reporting tools, and that distillation and a custom inference engine made the planning stage roughly ten times faster than the first internal version.

JJ Fiasson, Canva’s head of generative AI, framed the model’s purpose in a statement quoted by SmartCompany: “Craft delivers exceptional results for jobs where the layouts and text need to be perfect.” That is the use case the template-trained approach is meant to serve, and it explains why the announcement leans on legible text and editable layers instead of photographic realism. Images from Craft can be converted into editable layers with Canva’s Magic Layers feature.

SmartCompany also reports that owning the model lets Canva sell it to other businesses through the public API and an MCP server, which opens a revenue line beyond the editor. For free users the immediate fact is narrower: a new model is in the Dream Lab tab today, usable within a monthly allowance whose Craft-specific cost Canva has yet to state.

This article was produced with the assistance of AI and reviewed by Morning Overview editors prior to publication.

AI, explained every day: Get The Deep View →

We may earn a commission if you sign up.