Canva has rolled out an in-house AI model called Craft, built specifically to generate images inside its design platform. The move signals a shift away from leaning on third-party AI providers and toward owning more of the technology stack that powers its editing tools.
What happened
Craft is Canva's own image-generation model, developed internally rather than licensed from outside AI labs. Canva has previously relied on partnerships with companies like OpenAI and Google to power AI features inside its design tools, including text-to-image generation, background removal, and photo editing.
With Craft, the company says it can produce more polished, design-ready images with finer control over style, composition, and text rendering inside graphics — a notorious weak spot for many image generators, which tend to garble letters and logos. Canva has positioned Craft as tuned specifically for marketing and design use cases, rather than general-purpose image creation.
The model is being integrated into Canva's existing AI suite, known as Magic Studio, which already includes tools for resizing, background editing, and copywriting. Canva has not detailed separate pricing for Craft; it appears to be rolling out as part of existing subscription tiers rather than as a standalone paid add-on, at least for now.
Building an in-house model is a notable technical and financial undertaking. Training and running large image-generation models requires significant computing infrastructure, which is typically far more expensive than licensing access to someone else's model through an API.
Why it matters
Canva's move follows a broader pattern among software companies that started as design or productivity tools and have steadily built or acquired AI capabilities rather than permanently outsourcing them. Adobe has done something similar with its Firefly image model, built specifically to avoid copyright disputes tied to training data. Canva's approach appears aimed at a similar goal: tighter control over output quality, licensing terms, and long-term costs.
This also reflects a shift in how design software companies think about AI dependency. Relying entirely on outside model providers means being exposed to their price changes, usage limits, and policy shifts. Owning a model, even a narrower one built for specific tasks, gives a company more control over its own roadmap.
What this means for small businesses
For small business owners who already use Canva for social posts, ads, or marketing materials, Craft is likely to show up as better-looking AI-generated images with fewer obvious errors, particularly around text inside graphics. That's a meaningful quality-of-life improvement if you've ever tried to generate a promotional graphic only to get garbled lettering.
The bigger open question is cost and access. Companies that build their own AI infrastructure often start by including new features in existing plans, then later move premium capabilities behind higher-tier subscriptions once usage data shows where the demand is. Canva has done this before with other Magic Studio features, so it's worth watching whether Craft stays included in current plans or eventually becomes part of a pricier tier.
There's also a licensing question worth paying attention to. Businesses using AI-generated images in paid advertising or client-facing work should check Canva's terms around commercial use and copyright indemnification for Craft-generated content, since those terms can differ from general stock image licensing.
What to watch
Keep an eye on whether Canva introduces usage caps or a separate pricing tier for Craft in the coming months, and watch for how its text-rendering and brand-consistency features compare in practice to Adobe Firefly and other design-focused image models. Also worth tracking: whether Canva publishes details on what data trained Craft, which has been a point of scrutiny for other AI image tools.
The bottom line
Canva now controls more of its own AI pipeline instead of renting it from outside providers, which could mean steadier features and pricing — or set the stage for new tiers later. Businesses using Canva for marketing visuals should test Craft's output quality and review its commercial licensing terms before leaning on it for client or paid-ad work.