Canva unveiled Canva Craft on 22 October, the first image generation model built and trained from scratch by its in-house lab, Canva Research. The model works in two stages: it first plans the composition, meaning the elements, their positions and any text, and then renders the image at up to 2K resolution in more than 30 aspect ratios. Training placed heavy weight on millions of design templates from the platform, and it can render text in 55 languages. It is available to all Canva users and through a public API, with faster versions to follow.

What it means: Canva's move shows that large software platforms do not want to rely solely on third-party model providers for core features. With its own model, the company controls cost, speed and safety rules, and can tune it to design needs such as accurate text placement and editable layers. For Greek small businesses, communications agencies and online shops that already create posts and ads in Canva, the practical difference is that images should come out closer to finished material, with text that is legible and can be corrected. Multi-language support is interesting, although output quality in Greek should be tested before it is used in campaigns. The public API opens the way to building it into their own tools, for example e-commerce platforms that generate product images automatically. There are still open questions every business should check: the terms for commercial use, labelling of AI-generated content as the AI Act requires, and the risk of misleading product images, which can create problems under consumer protection law.

Read the original on Canva

Related service: Custom software

Related stories