Canva Craft Plans Designs Before Generating Images

Canva has introduced Canva Craft, its first in-house AI image generation model, and made it available to users. The launch marks a step deeper into foundational model development for a company best known for design software and templates. Craft uses a two-stage architecture that plans a design before generating an image, a process Canva says supports greater precision and customization. That planning step is the most distinctive part of the announcement. Many image generators respond directly to a prompt, which can produce attractive results but may not fit a specific layout, brand style, or design goal. Canva Craft appears designed to think about the composition and design intent before producing visuals. If it works as promised, users could get images that fit more naturally into presentations, social posts, marketing materials, and other projects. Craft also competes in a crowded field. OpenAI, Microsoft, Google, and other providers offer powerful image generation models, and many are improving quickly. Canva’s advantage is not necessarily raw model performance but distribution and workflow integration. Its tools are already used by millions of people who need to create designs without advanced technical skills. Embedding generation directly into that environment could make AI image creation feel like a natural part of the design process rather than a separate prompt-and-download step. The launch signals Canva’s ambition to control more of its AI stack. Relying on third-party models can be costly and limiting, while building in-house technology can allow tighter integration with templates, editing tools, and brand controls. It also creates a potential competitive moat if Craft can produce images that are easier to customize inside Canva than elsewhere. The challenge will be proving quality and reliability against well-funded rivals. Still, Canva Craft shows how design platforms are moving from adding AI features to building the models behind them.