Image models built around design constraints
Recraft develops generative image models alongside a design interface and editing tools. Its work pays attention to text size and placement, style references and repeatable visual direction, addressing the constraints of a designed asset rather than only producing a standalone illustration. [1]
The platform supports both raster imagery and vector artwork. Generating a scalable vector and converting an existing pixel image into a vector are separate operations; Recraft's API account describes both, along with background removal and upscaling. That makes its contribution useful in software that needs assets which can keep being edited or resized. [2]
The V3 release and developer access
Recraft V3 was announced on October 30, 2024. It added control over text positions and sizes, style creation and image editing. The authors' ranking and quality claims describe their launch evaluation, rather than a promise that every generated design meets a brief. [1]
A November 8 company post documented API availability and explicitly dated the model release to October 30. It described custom styles built from reference images and control over brand colors, bringing the model into automated design workflows. This dated account is more precise than assuming every capability on today's updated launch page arrived at the same moment. [2]
Sources
- Recraft introduces a revolutionary AI model that thinks in design language
Recraft · Oct 30, 2024
- Discover the Power of Recraft's image generation API
Recraft · Nov 8, 2024