Canva Launches Proprietary Canva Craft AI Image Generation Model

Industry Pulse News Desk · 2026-10-22

Visual communications platform Canva has introduced a two-stage artificial intelligence model designed to plan layout composition before rendering images.

Visual communications platform Canva unveiled its first proprietary artificial intelligence image generation model on Thursday, entering direct competition with major technology developers in the generative media space. The new system, designated Canva Craft, utilizes a two-stage operational framework designed to plan visual layouts before final rendering occurs.

Under the two-stage architecture, the model first constructs a structural layout plan before executing pixel-level image generation. The company stated the approach provides enhanced structural precision, visual consistency, and user control compared to standard single-pass text-to-image models.

Technical specifications released by the company indicate Canva Craft can produce visual assets at resolutions up to 2K across more than 30 distinct aspect ratios. The system is configured to generate a wide range of visual formats, including social media graphics, promotional materials, digital artwork, photorealistic scenes, and complex environmental imagery.

The deployment represents a strategic move by Canva toward developing in-house machine learning foundation models alongside its existing software ecosystem. The proprietary model positions the company to compete directly with generative image platforms operated by OpenAI, Microsoft, and Google.

The feature is being integrated directly into Canva’s existing suite of design tools, allowing individual and enterprise users to generate custom assets within active workspace projects.