Canva launches its first in-house image model

Canva has launched Craft, its first in-house image-generation model, making it available to users globally and through its public API. The company says the model was developed from scratch by its Canva Research team and can generate images at resolutions of up to 2K in more than 30 aspect ratios. Craft uses a two-stage architecture that separates planning a composition from rendering the image. Canva says the model first interprets a prompt and maps elements, spatial relationships, visual hierarchy and text, before a rendering model produces the image. The design is intended to improve object placement and text rendering and to let users adjust individual elements without substantially changing the rest of an image. The model is integrated into Canva’s Dream Lab and works with Magic Layers, which turns generated images into editable design elements, according to the company. Developers and product teams can also access Craft through the public API to add image generation to their own services. Canva has said an integration with its Model Context Protocol server is planned, but did not give a release date. Canva says Craft ranked fourth overall on DesignArena, a crowdsourced...