Canva Craft generates images from prompts

Canva has released Canva Craft, its first image generation model. Developed by the company’s research team, it is designed to compete with image models from Google, OpenAI, ByteDance and Meta.

The model works in two stages. First, it interprets a user’s prompt and creates a plan, including details such as the spatial relationships between elements, their relative scale and the image’s visual hierarchy. The plan can also include information for generating text. In the second stage, Canva Craft turns that plan into an image.

Canva says the model can produce images at up to 2K resolution in as many as 30 aspect ratios, to suit different platforms. Its two-stage approach is intended to help generate accurate object details and legible text. It also aims to let users make precise edits without changing other parts of an image.

Design training aims for accurate, editable results

Rather than training Canva Craft on photographs, Canva used human-designed templates from its library. These included posters, social posts, presentations, infographics, menus and typography-rich layouts. Canva said this design-focused training helps the model create images with accurate details and legible text, while allowing users to make precise edits without changing other parts of an image.

The model works in two stages: it first interprets a prompt and plans details such as spatial relationships, relative scale and visual hierarchy, along with information for text generation. It then generates an image of up to 2K resolution in as many as 30 aspect ratios for different platforms.

Generated images can also become layered designs through Canva Magic, the company’s image-to-design transformer. Canva released its own model that understands designs and layers last year, enabling users to take an image created with Craft and turn it into an editable design.

Availability, integrations and model development

Canva Craft is available to all users on Canva’s platform through the Dream Lab tab, and the company says the model will also be available through the Canva API. Canva plans to bring image generation to third-party chatbots, including ChatGPT and Gemini, through Canva’s MCP.

Researchers from Leonardo.ai contributed to the model. Canva acquired the media generation platform in 2024, and says its researchers helped develop Craft. The company is also working on two cost-efficient versions, Craftfast and Craft-ultrafast, which it plans to release soon.

Canva has used a range of image models for generation on its platform. It said Craft could help reduce API costs, while also giving the platform greater control over quality and speed. The company’s stated benefits encompass both the results users receive and the resources required to provide image generation at scale.

Canva’s business context

Canva’s first image-generation model arrives as the design platform develops its own AI capabilities and manages expectations for future growth. In 2024, the company acquired media-generation platform Leonardo.ai, whose researchers contributed to Canva Craft.

The model also sits within a business that was on track for $4 billion in revenue the previous year. But Canva cut its growth forecast for 2026, and external analysts reportedly lowered their valuation of the company from $42 billion to $35 billion.

Canva has used a range of image-generation models on its platform, and said its own model gives it greater control over quality and speed, in addition to helping with costs. Building on research that includes Leonardo’s contribution, the company is developing two cost-efficient versions, Craftfast and Craft-ultrafast, which it said would be released soon.