Recraft releases V4.1 image model
Recraft has released V4.1, a point upgrade to its image-generation model with Vector and Utility variants, and says it produces more natural photos and better aesthetic results from short prompts.
Recraft announced V4.1, an update to its image-generation model. The company says it improves photorealism, adds illustration styles, and handles short prompts better.
The release has three variants:
- V4.1, the main model, which Recraft calls its "most expressive." It is built for exploration and concepting and makes its own choices about light, mood and composition.
- V4.1 Vector, built for logos, typography and illustration, where the company stresses lettering and precise form.
- V4.1 Utility, built for predictable output. It uses flat lighting and simpler compositions for projects that "don't need drama."
Recraft's case rests on side-by-side images against GPT Image 2, Midjourney and Nano Banana 2. The company says V4.1 produces calmer backgrounds and more candid-looking people. It says GPT Image 2 output looks more like stock photography. Recraft also says V4.1 can get a style right from prompts of three to five words, while rivals add clutter, such as extra neon signs from Nano Banana 2. Other examples cover 3D chrome renders, gradients, logo lettering and a character reference sheet with front, profile, back and 3/4 views.
The post publishes no benchmarks, preference-test results or independent evaluations. The comparisons are a small set of images chosen by Recraft. The announcement also leaves out pricing, API availability and technical details such as architecture or training changes.