Canva’s first AI image model plans designs before generating them

Live updates

1. 45m ago

   Canva Launches Craft In-House AI Image Generation Model

   Canva has released its first in-house image generation model, named Canva Craft, designed to compete with major rivals like OpenAI, Google, Meta, and ByteDance. Launched on 22 October 2026, the tool employs a unique two-stage architecture that plans layouts and text before generating graphics at up to 2K resolution. The model is available to all users and developers through Dream Lab, targeting posters and social media graphics with editable, layer-by-layer outputs. Industry testing indicates mixed results, while benchmark evaluations rank the model fourth on DesignArena behind three OpenAI models.

   Why it matters

   The introduction of Canva Craft marks a strategic push for the design platform as it works to retain customers migrating to competitor AI tools. Built from scratch and trained on human-designed templates rather than photographs, the model addresses common industry concerns regarding training data and text rendering. This launch expands Canva's generative capabilities as the company broadens its automated design suite.

   What is confirmed

   • Canva launched its first in-house image generation model, Canva Craft, on 22 October 2026.
   • Canva Craft utilizes a two-stage process that plans the layout and words before drawing images at up to 2K resolution.
   • The model was trained on human-designed templates and was not trained on photographs.
   • Canva Craft is accessible to all Canva users through Dream Lab and supports over 30 aspect ratios.
   • The model ranks fourth on DesignArena behind three OpenAI models.

   Still unconfirmed

   • The AI image generator delivers mixed performance results when put through paces by reviewers.

   What to watch next

   • Broader adoption metrics and performance comparisons against competing tools like Midjourney and ChatGPT
   • Developer integration updates and user reception of layer-by-layer editable graphics
   Sources used for this update (16)

   1. docs.lovable.dev — Lovable changelog - Lovable Documentation
   2. www.cnet.com — iPhone 18: What We Know About Apple's Spring Lineup
   3. www.theverge.com — Here’s Nintendo’s overview trailer for the Ocarina of Time remake.
   4. The Verge — Canva’s first AI image model plans designs before generating them
   5. TechCrunch — Canva releases its first image generation model
   6. Business Insider — I put Canva's new AI image generator through its paces. The results were a mixed bag.
   7. Rampart — Canva’s AI transformation comes with a cost
   8. PetaPixel — Canva Releases AI Image Generator That Wasn’t Trained on Photographs
   9. www.startupdaily.net — Canva launches a home-grown AI image generation model - Startup Daily
   10. cryptobriefing.com — Canva releases Canva Craft, its first in-house image generation model - Crypto Briefing
   11. gadgetsnow.indiatimes.com — Canva Craft Plans Before It Paints: ChatGPT, Nano Banana And Midjourney Compared
   12. www.ilearnlot.com — How To Create Realistic AI Image Generator Canva - Ilearnlot
   confidence 95%