Canva launches first AI image model with planning for precise designs

Canva has launched its first in-house AI image generation model named Canva Craft, according to The Verge. The model can create images at resolutions up to 2K and supports over 30 aspect ratios, aiming to enhance precision and consistency in design outputs.

Unlike traditional models, Canva Craft uses a two-stage architecture that first plans the design by interpreting user prompts and detailing spatial relationships, scale, and text elements before generating the final image, TechCrunch reported. This method allows for more accurate and legible images.

The model was trained on millions of human-made templates from Canva's extensive content library, including posters, social media posts, presentations, and typography-focused layouts. This training approach helps Canva Craft produce images that can be converted into layered designs using Canva Magic, the company’s image-to-design transformer.

Canva Craft is integrated into the Canva platform and accessible to all users via the Dream Lab tab. Additionally, it will be available through Canva's API and is planned to support image generation in third-party chatbots such as ChatGPT and Gemini through Canva’s MCP, according to TechCrunch.

The new model is expected to improve image quality, speed, and cost efficiency for Canva, building on the company’s prior AI developments like prompt-based editing and layered AI-generated images, as detailed by The Verge.