Hajra Naz
Design platform Canva released its first image generation model. The model is called Canva Craft. The company’s research team developed the software. It aims to compete with similar models from tech giants like Google, OpenAI, ByteDance, and Meta.
The company explained that its model operates in two distinct stages. During the first stage, the system interprets user prompts. It then builds a layout plan. This plan includes object placement, relative scale, and visual hierarchy. It also includes specific text generation data. During the second stage, the model renders the image. It supports resolutions up to 2K. It also provides up to thirty different aspect ratios for various digital platforms.
Canva stated that this multi-stage approach helps produce accurate object details. It also ensures that generated text remains clear and legible. Furthermore, users can make precise edits. They can alter specific sections without disrupting other parts of the image.
The company noted a unique training approach. Instead of relying purely on raw photographs, the model was trained using human-designed templates from Canva’s library. These training materials included posters, social media graphics, presentations, infographics, menus, and typography-focused layouts.
Read More: Canva Unveils Its Own Design Model and Powerful New AI Tools
As a result, images created with Canva Craft integrate with other tools. Users can convert the generated visuals into fully layered designs using Canva Magic. That feature utilizes the company’s proprietary image-to-design transformer. Last year, the platform launched its own foundational model capable of understanding complex design layers.
Canva Craft is now live for all platform users. Customers can access the model directly through the Dream Lab tab. The model is also accessible via the Canva API. Additionally, the company plans to enable image generation inside third-party chatbots like ChatGPT and Gemini through Canva’s MCP.
Canva has integrated various third-party image models in the past. Developing an in-house model will likely reduce external API expenses for the company. Executives stated that beyond cost savings, the proprietary model gives the platform better control over generation speed and output quality.
In 2024, Canva acquired Leonardo.ai, a specialized media generation platform. Company leaders confirmed that researchers from Leonardo contributed heavily to the development of this new image model.
Read More: Canva is Now Inside ChatGPT – Create Logos, Presentations, and Social Media Posts Instantly
Canva announced that it is currently building two additional versions. These upcoming variants are named Craftfast and Craft-ultrafast. The company plans to release them in the near future.
Financial reports indicate shifting market conditions for the design giant. Canva was on track to achieve $4 billion in revenue last year. However, the company adjusted its growth forecast for 2026. External analysts report that these changes caused its private valuation to drop from $42 billion down to $35 billion.
