Canva releases its first image generation model

Canva, the popular design platform, has launched its inaugural image generation model, named Canva Craft. Created by the company's research team, this model aims to compete with similar models from industry giants such as Google, OpenAI, ByteDance, and Meta. Canva Craft operates in two stages: initially, it interprets users' prompts and crafts a plan encompassing elements like spatial relationships, relative scale, visual hierarchy, and text generation guidelines; subsequently, it generates an image, up to 2K resolution in any of 30 aspect ratios suited for various platforms.

This method ensures precise details of objects and legible text, allowing users to make exact edits without altering other image sections. Instead of employing photographs, Canva trained this model using human-created templates from its library, including posters, social media posts, presentations, infographics, menus, and typography-heavy layouts.

Consequently, Canva Craft-generated images can be transformed into layered designs using Canva Magic, the platform's image-to-design transformer. Introduced last year, Canva Craft is accessible to all users via Canva's platform, found in the Dream Lab tab, and also through the Canva API. The company intends to extend image generation support to third-party chatbots like ChatGPT and Gemini via Canva's Model Customization Protocol (MCP).

Given its previous use of several image generation models, Canva believes these new models can help reduce API expenses while offering superior quality and speed control. In 2024, Canva acquired media generation platform Leonardo.ai, and researchers from the company contributed to the development of this new image model. While Canva projected $4 billion in revenue last year, it revised its growth forecast for 2026 downwards, leading to a drop in the company's valuation from $42 billion to $35 billion, according to external analysts.