AI
Canva Launches Craft, Its First In-House AI Image Model Trained on Designs Instead of Photographs
Canva launched its first in-house AI image generation model, called Craft, on October 22, 2026. The model was developed by Canva's research team and trained on millions of ready-made design templates — such as posters, menus, and presentations — rather than photographs. It became immediately available to all users through the Dream Lab tool, and to developers through the Connect API under a new model name, craft_v1_2k.
Craft works in two stages: the first interprets the text prompt and drafts a plan for visual hierarchy, element placement, scale, and typography, and the second actually renders the image based on that plan. The model outputs images up to 2K resolution in more than 30 aspect ratios, and supports rendering text within images in 55 languages.
Canva says faster versions of the model are planned, along with an MCP server integration for AI assistants, though no timeline has been given for either.
Source: PetaPixel
Need filming at the level you just read about? I reply within 24 hours.
Start your project