Canva Unveils In-House AI Image Model Called Canva Craft
Canva is expanding its artificial intelligence capabilities with the introduction of its first in-house image generation model, designed to compete directly with industry offerings from OpenAI, Microsoft, and Google.
Radwan AbdulhadiEditor and technology writer
Introduction to Canva Craft
Canva is pushing deeper into the artificial intelligence landscape with the release of an image generation model that goes head-to-head with models from Google, Microsoft, and OpenAI. Revealed through [an announcement on Thursday](https://go.skimresources.com/?id=87443X1797501&xs=1&url=https%3A%2F%2Fwww.canva.com%2Fnewsroom%2Fnews%2Fcanva-craft%2F), the new system is [called Canva Craft](https://go.skimresources.com/?id=87443X1797501&xs=1&url=https%3A%2F%2Fwww.canva.com%2Fdream-lab) and introduces a novel approach to automated design creation.
Unlike standard text-to-image generators that immediately render pixels from a prompt, Canva Craft utilizes a specialized two-stage architecture. According to the company, this setup allows the software to plan out a design before actually generating the final image. The result, the company claims, is significantly greater precision, consistency, and overall control for creators.
Capabilities and Technical Specifications
In terms of output quality and formats, Canva Craft is capable of generating images at up to 2K resolution. Furthermore, users can select from more than 30 aspect ratios to suit their specific projects.
The platform notes that the technology is engineered to handle a broad range of visual tasks. Beyond generating routine social media posts and standard marketing content, Canva Craft is intended to create realistic scenes, digital art, environmental imagery, and other complex visual assets.
Canva highlights that the model's two-stage system is particularly useful for tackling complex visual challenges that traditionally trip up AI systems, such as rendering legible text within an image and executing precise object placement across the canvas.
Building on Previous AI Integration Efforts
The introduction of Canva Craft builds heavily on the platform's recent comprehensive push into artificial intelligence tools. Earlier in the year, [Canva launched new prompt-based editing capabilities](https://www.theverge.com/tech/913068/canva-ai-2-update-prompt-based-editing-availability) that enabled users to create and modify designs simply by describing their intent in natural language.
Additionally, the design software [also brought layers to AI-generated images](https://www.theverge.com/tech/893124/canva-ai-magic-layers-feature-beta), giving creators the ability to isolate and alter specific parts of an image independently while keeping the rest of the composition intact.
Training Data and Benchmark Rankings
To achieve its familiarity with practical design needs, Canva trained Craft on millions of human-made templates sourced directly from the platform's own extensive content library. This training corpus included presentations, social media layouts, infographics, menus, and various other professional templates.
Early performance indicators suggest the model holds its own against established competitors. On [Design Arena’s leaderboard](https://www.designarena.ai/leaderboard?tab=image§ion=image), which serves as a crowdsourced benchmark for evaluating AI image generation models, Canva Craft currently holds the fourth position.
On this crowdsourced leaderboard, Canva Craft ranks just behind OpenAI's trio of competing models: GPT-Image-2.5 Sunburst, GPT-Image-2.5 Flare, and GPT-Image-2.
Safety, Availability, and Future Plans
Alongside its creative features, Canva emphasizes that the model is equipped with robust input and output moderation systems. It also includes built-in privacy and safety controls alongside dedicated reporting tools to manage potential misuse.
Looking ahead, Canva intends to expand the ecosystem surrounding the technology. The company plans to allow external AI agents to integrate with Canva Craft using the open Model Context Protocol. Furthermore, future updates will introduce more efficient variations of the model, specifically designated as Craft-fast and Craft-ultrafast.
Canva Craft is available globally to all users starting immediately. Those interested in testing out the new image model can access it by navigating to the three-dot option located on the left-hand menu of the Canva interface and then selecting Dream Lab.
Sources
- The VergeCanva’s first AI image model plans designs before generating them
Continue chronologically
Related entity coverage
Related stories
Google, OpenAI, Anthropic Join New Scraping Controls
Google, OpenAI, Anthropic, and Microsoft pledge to support a new common robots.txt signal to block web scraping for generative AI training.
Major AI firms back new multi-company web scraping opt-out standard
OpenAI, Google, Microsoft, and Anthropic plan to adopt a new shared policy token standard allowing publishers to manage AI training preferences collectively.
Google's SynthID AI Detector Goes Public for Everyone
Google's SynthID Detector is now available globally for regular users, offering a simple way to check if images, videos, or audio files are AI-generated.
Node2Shell Brotli Flaw Triggers Industry Patches
Critical Node2Shell Brotli decoder vulnerability leads to security updates across Microsoft, Node.js, and Python packages.
