Recraft V4 Styles Adds Reference-Led Generation
The design model can derive a reusable visual style from one to ten reference images without requiring a separate training process.
Style extraction without training
Recraft launched V4 Styles, a reference-led image-generation mode intended to preserve a visual identity across multiple assets. A user can supply a single reference image in Recraft Studio, while API integrations support between one and ten references. The model then applies the inferred style to subsequent generations without requiring fine-tuning or a separately trained adapter.
The feature produces raster images and editable SVG vectors, extending style control beyond conventional text-to-image output. Recraft says the reference is used to guide characteristics such as texture, color treatment and graphic language without copying its depicted subject. V4 Styles was released through Recraft’s own product and distributed at launch through partners including OpenRouter, Replicate and fal.
A production workflow, not a one-off prompt
Consistent style has remained a practical weakness of generative design systems. A model may create a strong individual image but drift when asked to build a campaign, icon family or set of illustrations. Recraft’s approach turns the inferred style into a reusable production control, allowing teams to generate multiple assets without describing the same visual rules repeatedly.
Support for up to ten API references also gives developers a way to represent a broader brand vocabulary than a single image can carry. However, the launch materials do not provide a public quantitative measure of style fidelity or subject leakage. The absence of training lowers setup time, but it also leaves open how reliably the system separates style from composition, recognizable characters or protected creative elements.
Why it matters
Generative-image competition is shifting from isolated image quality toward repeatability and editability. Recraft is targeting professional design work where a coherent series is more valuable than a striking standalone result, and native SVG output makes the generated material easier to revise in existing production pipelines. The feature’s commercial significance will depend on whether it maintains consistency across varied subjects without reproducing unwanted details from references. That unresolved boundary is especially important for agencies and brands using third-party artwork as inspiration.