Design platform Canva has introduced its first image generation model, called Canva Craft, as it expands its artificial intelligence capabilities and competes with image generation systems developed by Google, OpenAI, ByteDance and Meta.

The model was developed by Canva’s research team and is designed to produce detailed images with accurate objects, readable text and greater control over editing.

The company said the system uses a two stage process, beginning with an interpretation of the user’s prompt before developing a plan that considers spatial relationships, the relative size of objects, visual hierarchy and text placement.

During the second stage, Canva Craft generates images at resolutions of up to 2K and supports approximately 30 aspect ratios for different platforms and purposes.

According to Canva, the approach helps the model create more accurate visual compositions and legible text, addressing common challenges associated with AI generated images.

The company also said users can make precise changes to selected parts of an image without unintentionally altering other elements.

Unlike image generation models trained primarily on photographs, Canva Craft was trained using design materials from Canva’s library.

These include posters, social media graphics, presentations, infographics, menus and layouts featuring substantial amounts of text.

Canva believes this training approach will make the model particularly useful for users who need images that can be incorporated into practical design projects rather than standalone visual content.

The model is available through Canva’s platform, where users can access it through the Dream Lab section.

It is also expected to be accessible through the Canva application programming interface, while the company plans to support image generation through third party chatbots such as ChatGPT and Gemini using its Model Context Protocol integration.

Canva said developing its own image generation technology could reduce its dependence on external models and lower application programming interface costs, while giving it greater control over output quality and generation speed.

Canva acquired image generation company Leonardo.ai in 2024, and researchers from Leonardo contributed to the development of Canva Craft.

The company is also developing two faster and more cost efficient versions, called Craftfast and Craft ultrafast, which are expected to be released later. The launch comes as Canva continues to expand its AI offerings amid changing business expectations.

The company was reportedly on track to generate $4 billion in revenue last year, but it reduced its growth forecast for 2026, contributing to reports that its valuation had fallen from $42 billion to $35 billion, according to external analysts.

Source: TechCrunch