Recraft: Recraft V4 Styles Vector

Recraft V4 Styles Vector is an image generation model from Recraft, listed on api14 under the model id recraft/recraft-v4-styles-vector. It sits in the image category and is built around style consistency: rather than generating an image from a text prompt alone, it takes style reference images as part of every request and produces a new image that reproduces the visual style of those references. Recraft is the provider behind the model, and the provider release date is 2026-08-26.

The model accepts text and image inputs and returns images as its only output. Its context window is 65,536 tokens, with a maximum output of 58,982. The image parameters define the shape of a request: aspect_ratio is an enum covering 1:1, 2:1, 1:2, 3:2, 2:3, 4:3, 3:4, 5:4, 4:5, 16:9, 9:16 and auto; output_format is an enum whose only listed value is svg; n is a range from 1 to 6; and input_references is a range from 1 to 10. Because output_format is fixed to svg, the model produces vector output rather than raster images.

The combination of style references and SVG output points to work where artwork has to match an existing visual language and stay editable afterwards. Teams can supply between one and ten reference images to steer the look, then generate up to six variations per request to explore options. Typical fits include icon sets and UI glyphs, logos and brand marks, illustrations for documentation or marketing, and diagram or infographic assets that need to scale cleanly. Because the output is vector, the results can be recoloured, resized and adjusted in design tools without the loss of quality that comes with scaling raster images.

You call the model through api14's OpenAI-compatible API using the model id recraft/recraft-v4-styles-vector. Requests follow the same shape as other OpenAI-compatible endpoints, with your text prompt and reference images as input, and the image parameters passed alongside. Note that supported_parameters is empty, so the tunable surface is the image parameter set described above: aspect_ratio, output_format, n and input_references. Every request needs at least one style reference image, so plan for reference assets before you start integrating.

That requirement is also the main constraint. If you want to generate images purely from a text description with no style reference, this model is not the right fit, since at least one reference image is mandatory. Similarly, if you need raster output formats such as PNG or JPEG, the only listed output_format here is svg. For style-driven vector work, however, the model gives you a direct path from reference images to scalable artwork.