Canva Launches Craft Image Model With Design-Focused Generation and Editing

TL;DR

• Canva has introduced Canva Craft, its first image generation model.
• The model plans layouts before generating images at resolutions up to 2K.
• It supports 30 aspect ratios and is designed to improve text accuracy and precise editing.
• Craft is available through Dream Lab, with API access and third-party chatbot integrations planned.

Canva Craft Plans Designs Before Generating Images

Canva has released its first image generation model, Canva Craft, expanding its in-house AI capabilities as it competes with offerings from Google, OpenAI, ByteDance and Meta.

Developed by Canva’s research team, Craft uses a two-stage process intended to improve how generated images handle composition, objects and text.

First, the model interprets a user’s prompt and develops a plan covering spatial relationships, relative scale, visual hierarchy and text requirements. It then generates the image, supporting resolutions up to 2K and 30 aspect ratios suited to different publishing formats.

Canva said the approach helps produce accurate object details and legible text. The company also said users can make targeted edits while preserving other parts of an image.

Training Focuses on Templates and Typography

Canva said it trained Craft using designs rather than photographs, drawing on human-created templates from its library.

The training material includes posters, social media graphics, presentations, infographics, menus and layouts containing substantial typography.

That focus positions Craft around the structured compositions commonly created on Canva, where text placement and visual hierarchy are central to the finished product.

Images generated with Craft can also be converted into layered designs through Canva Magic, the company’s image-to-design transformer. This builds on the design model Canva introduced last year to understand layouts and layers.

Dream Lab Access Arrives With Broader Integrations Planned

Canva Craft is available to all users through the platform’s Dream Lab tab.

The company said it will also offer the model through the Canva API. Planned support through Canva’s Model Context Protocol integration would enable image generation within third-party chatbots such as ChatGPT and Gemini.

The rollout extends Canva’s effort to make its design tools accessible both within its own platform and through external AI services.

Researchers from Leonardo.ai, the media generation platform Canva acquired in 2024, contributed to Craft’s development.

Canva Targets Greater Control Over AI Costs and Performance

Canva has previously integrated multiple external image generation models into its platform. Developing its own model could reduce reliance on third-party APIs while giving it greater control over performance.

The company said its objectives extend beyond cost savings to include better oversight of generation quality and speed.

Two additional versions, Craftfast and Craft-ultrafast, are in development and are expected to launch soon. Canva described them as more cost-efficient alternatives but did not provide release dates.

The launch comes as the company faces pressure over AI spending and growth expectations. Canva was on track for $4 billion in revenue last year, but subsequently lowered its 2026 growth forecast. External analysts have linked that revision to a reported valuation decline from $42 billion to $35 billion.