Back to all articles
•Technology

Canva Launches Canva Craft: New Image Generation Model Shakes Up AI Design Sector

View original source

Canva has unveiled its first image generation model, Canva Craft, aimed to rival offerings from tech giants like Google, OpenAI, ByteDance, and Meta. This model, developed by Canva's own research team, operates in two stages:

  • Stage 1: Interprets users’ prompts and develops a plan incorporating spatial relationships, scales, visual hierarchy, and text generation details.
  • Stage 2: Generates images up to 2K resolution, accommodating 30 different aspect ratios suited for various platforms.

Key Features:

  • Enhanced detail accuracy and text legibility.
  • Ability to precisely edit images without altering unrelated sections.
  • Training harnessed directly from Canva's template library, not photographs, spanning across posters, social media posts, presentations, infographics, and more.

Integrated with Canva Magic, these images can transform into layered designs. The introduction of Canva Craft is accessible through the Dream Lab tab, with plans for wider availability via Canva's API and integration into chatbots like ChatGPT and Gemini.

The strategy is designed to minimize API expenses while boosting quality and speed control. Canva also plans to introduce two efficient versions, Craftfast and Craft-ultrafast.

The launch follows Canva's acquisition of Leonardo.ai in 2024, leveraging talent and technology from this media generation platform. However, Canva has adjusted its ambitious revenue targets for 2026, dropping its valuation from $42 billion to $35 billion.