The model supports resolutions up to 2K and offers over 30 aspect ratios, targeting both professional marketing assets and creative digital art. Unlike standard diffusion models, Craft was trained on millions of human-made templates from Canva’s own library, including infographics, menus, and presentation slides. This training focus aims to solve persistent industry issues regarding complex layouts and accurate text rendering within generated imagery.
Canva launches Craft AI to challenge OpenAI and Google image models
Canva has introduced Craft, a proprietary image generation model designed to rival top-tier competitors by prioritizing structural planning over raw output. By utilizing a two-stage architecture, the system maps out design compositions before rendering, promising users heightened precision in object placement and text legibility across diverse visual formats.
Currently holding the fourth position on the Design Arena leaderboard, Craft trails three iterations of OpenAI’s image models. To bridge the gap, the company has integrated robust privacy, safety, and content moderation tools into the interface. Users can access the feature immediately via the Dream Lab option located in the platform's sidebar. Future plans include the release of accelerated Craft-fast and Craft-ultrafast variants, alongside an open Model Context Protocol to facilitate third-party AI agent integration.
Comments (0)
No comments yet. Be the first!