News · Images

Canva launches Canva Craft, its first image generation model

Canva Craft is rolling out to all Canva users with design-focused image generation, layered editing potential and API access.

By Emonarc Editorial Team4 min read
A designer’s hands assemble one image into layered paper mockups for a post, poster, and presentation card on a dark table.

Canva has released Canva Craft, its first image generation model, according to TechCrunch. The launch matters because the model is built around design work rather than generic image creation, giving Canva users a more native way to make visuals for social posts, presentations, menus and other layout-heavy assets.

The model is available to all users inside Canva through the Dream Lab tab, and Canva also plans to offer it through its API. For creators, marketers and small businesses already working in Canva, the practical question is whether this can reduce the need to jump between a separate image generator and a design editor.

Canva Craft is tuned for design, not just pictures

Canva said Canva Craft works in two phases. First, it reads the user’s prompt and builds a plan for the image, including how elements should sit in relation to each other, their relative size, visual hierarchy and details needed for text generation. Then it creates the image, with output up to 2K resolution and support for 30 possible aspect ratios.

That design-first planning step is the key distinction Canva is emphasizing. The company says the approach helps produce more accurate object details and readable text, while also allowing users to make targeted edits without changing unrelated parts of the image.

For everyday Canva users, that could be useful for common tasks where layout and wording matter: event posters, product announcements, slide visuals, restaurant menus, infographics and social media graphics. Those are also the kinds of assets Canva says were part of the training mix.

Trained on Canva-style templates

Canva said it trained the model on human-designed templates from its own library rather than on photographs. The source material included posters, social posts, presentations, infographics, menus and typography-heavy layouts.

That choice explains why Canva is positioning Craft as more than a prompt-to-image feature. If the model understands design conventions, it may be better suited to producing assets that are meant to be edited, resized and reused inside Canva, instead of static images that need to be rebuilt elsewhere.

The model can also connect with Canva Magic, the company’s image-to-design transformer. According to Canva, images made with Canva Craft can be converted into layered designs through that system. Canva previously released a model that understands designs and layers, which gives the new image generator a path into editable Canva workflows.

For small teams, this is the important workflow shift: the output is not only something to download. It can become a design object that can be adjusted, rearranged and incorporated into an existing Canva project.

Lower platform costs, but no user pricing details

Canva has already used several image generation models inside its platform, according to TechCrunch. Bringing its own model into the stack could help the company reduce API costs, and Canva said the model also gives it more control over quality and speed.

That does not mean Canva announced a new price for users. The source does not state any change to subscription plans, credit systems or free-tier limits tied to Canva Craft. For customers, the immediate benefit is availability inside the product, not a confirmed reduction in what they pay.

The company is also working on two cost-efficient versions of the model, called Craftfast and Craft-ultrafast, which it says are coming soon. Canva has not provided further launch timing in the source material.

API and chatbot plans could widen its reach

Canva Craft is not limited to the web interface. Canva said the model will be available through the Canva API, which could matter for businesses that generate branded or repeatable design assets through their own systems.

The company also plans to support image generation in third-party chatbots such as ChatGPT and Google Gemini through Canva’s MCP. That would put Canva-generated design assets closer to the places where many users already draft campaigns, brainstorm creative ideas or plan content.

Canva is entering a crowded market. TechCrunch notes that the company is competing with image generation models from Google, OpenAI, ByteDance and Meta. But Canva’s advantage is less about being a standalone image model and more about sitting inside a design platform where users can immediately turn generated assets into posts, presentations and other finished materials.

The Leonardo.ai connection is also notable. Canva acquired Leonardo AI in 2024, and the company said researchers from Leonardo contributed to Canva Craft. That gives Canva an internal media-generation team while keeping the model closely tied to Canva’s design product.

For creators and marketers, the takeaway is straightforward: Canva Craft is worth trying first when the end product is a Canva design, especially if text, layout and aspect ratio matter. Dedicated image generators may still be part of some workflows, but Canva is trying to make the default option the one already built into the design canvas.

Frequently asked questions

What is Canva Craft?

Canva Craft is Canva’s first image generation model, developed by the company’s research team.

Where can Canva users access Canva Craft?

Canva says all users can access Canva Craft from the Dream Lab tab on its platform.

Will Canva Craft be available outside Canva’s main app?

Yes. Canva says the model will be available through the Canva API and plans support for third-party chatbots like ChatGPT and Gemini through Canva’s MCP.

Sources

  1. TechCrunch: Canva releases its first image generation model

The daily AI brief. 5 minutes, free.

What shipped, what changed in pricing, and the tools actually worth paying for.

No spam. Unsubscribe in one click.