Canva Craft: What Canva’s AI Image Model Means for Business

Canva Craft is Canva’s first image generation model built and trained from scratch by its own AI lab, and it is now available to every Canva user and to developers through Canva’s public API. For Australian businesses that already make their social posts, flyers and presentations in Canva, it means Canva’s AI image generator is now powered by a model trained by the Australian company itself, with an API that lets you build the same model into your own website, store or marketing system. Here is what Canva announced, where Craft fits in a small business, and what to check before you use it commercially.

What Canva announced

Canva unveiled Canva Craft on 22 October (US time), describing it as the first foundation model built and trained from scratch by Canva Research, its in-house AI lab. It generates images at up to 2K resolution in more than 30 aspect ratios, takes prompts in any language and can render text inside images in 55 languages.

The interesting part is how it works. Most image models go straight from a prompt to pixels. Craft uses two stages: a planning model first writes out the composition (which elements appear, where they sit, how big they are, the visual hierarchy and any text), then a separate rendering model turns that plan into the image. Canva says this gives more legible text, more precise placement and the ability to change one element without disturbing the rest.

Canva also says Craft was trained with a much greater emphasis on design than on photographs, using human-made templates from its own content library, such as posters, social posts, presentations, infographics and menus, then refined with feedback from Canva’s professional designers. TechCrunch reports that researchers from Leonardo.ai, which Canva acquired in 2024, contributed to the model, and that Canva is using it to compete with image models from Google, OpenAI, ByteDance and Meta. Canva points to a strong debut on the DesignArena leaderboard, a crowdsourced comparison of AI design output, and says faster, cheaper variants called Craft-fast and Craft-ultrafast are on the way.

Why an Australian-built image model matters

Until now, Canva has relied on a mix of image models in its products. As SmartCompany points out, owning the model means Canva can now sell it to other brands through its API and MCP server. For local businesses, the practical effect is simpler: the AI image tool your marketing team already has open is now one of the serious contenders, and it was built with design layouts in mind rather than just photography.

That matters because a lot of everyday small business imagery is not photography. It is a menu board, an event flyer, a sale banner, a slide or a social tile with words on it, which is exactly where earlier image generators tended to mangle the text.

Where Canva Craft fits in a small business

Craft is available in Dream Lab inside Canva. Canva says it is aimed at marketing assets, editorial-style photography, illustration, portraits, landscapes and conceptual work. Sensible first uses include:

• Social posts and promotions with headline text that lands where you want it, in the sizes each platform needs.
• Flyers, menus and posters where layout and hierarchy matter, which can then be turned into editable layers with Canva’s Magic Layers.
• Multilingual material for customers who read languages other than English, checked by a fluent speaker before it goes out.
• Background and mood imagery for presentations, landing pages and proposals.

Be realistic about the limits. Canva itself says it is still working on fine print and technical drawings, so do not rely on Craft for anything where small details must be exactly right, such as safety signage, product specifications or prices. A person should still read every word before it is published.

Building Canva Craft into your own tools

For businesses with their own software, the bigger news is the API. Canva says developers and product teams can build Craft into marketing platforms, e-commerce sites, ad builders and print services. Its developer announcement and the API reference set out how it works, and a few details matter when you are planning a project:

• It is a preview API. Canva warns there may be unannounced breaking changes, and public apps that use preview APIs will not pass Canva’s app review. That suits internal tools and pilots better than a product you sell to others.
• It acts on behalf of a Canva user. Requests need a user access token, and each image uses that user’s AI credit allowance, so plan whose account and credits a workflow runs on.
• It is not instant. Because Craft plans before it renders, Canva says a job takes roughly 50 to 60 seconds, so your system should queue the request and check back rather than make a customer wait on a spinning page.
• Plan for failure. If you pin the Craft model and it is unavailable, the job fails rather than silently switching models. Rejected prompts come back as unsafe input. Download links expire after 24 hours, so save the images you keep.

Canva also says Craft is coming to the Canva MCP server, so AI agents and assistants can create design-ready images as one step in a bigger task, such as drafting a launch graphic alongside the announcement copy. If you are connecting AI models to your own systems, our LLM integration and custom AI development work covers queues, fallbacks, credentials and approval steps like these.

A Canva Craft checklist before commercial use

1. Read the ownership terms. Canva’s AI Product Terms say that, as between you and Canva, you own your output, except where it modifies or incorporates Canva library content, which stays licensed to you rather than owned.
2. Do not use it for anything you need to own exclusively. The same terms note that outputs may not be unique and other users may receive similar results. That makes generated images a poor choice for a logo or anything you might want to register as a trade mark.
3. Be honest about AI images. Canva’s terms prohibit using its AI to mislead anyone that content is human-made, and prohibit removing provenance or metadata tags such as C2PA. Make sure your workflow does not strip them, and never let a generated “product shot” show something your actual product is not.
4. Check your privacy settings. Canva says your privacy settings control whether your data can be used to improve its AI services, so decide what you want before staff upload client material.
5. Name an approver. Pick the person who checks AI images for accuracy, brand fit and claims before they go live. Our earlier piece on AI copyright in Australia covers the wider questions about AI tools and content rights.

What this means for your marketing and software plans

Canva Craft lowers the effort of making on-brand visuals, but the business questions stay the same: which tasks are worth automating, who signs off, and what happens when a tool changes. If your team is new to this, our free AI marketing course covers approved content and honest advertising, and the AI prompts course covers writing and testing prompts. For businesses weighing up image generation inside their own website, store or CRM, an AI business audit maps where it would actually save time; here is what an AI audit includes.

This is general information based on Canva’s announcement, developer documentation and AI Product Terms, and published media reporting as at 26 October 2026. It is not legal advice. Terms and API details can change, so check Canva’s current terms and get advice on your own situation before relying on AI-generated images commercially.

Related reading: Canva Craft sits in the design and marketing row of our jobs-based AI tools guide for small business, alongside the other marketing and admin tools Australian teams actually use. Try the free Start Smarter course if you are choosing a first AI tool.

Featured image: Photo by Jakub Zerdzicki on Pexels.