Recraft: Recraft V4 Styles Pro Vector

Recraft V4 Styles Pro Vector is an image generation model from Recraft, listed on api14 under the model id recraft/recraft-v4-styles-pro-vector. It sits in the image category and is built specifically for style-consistent vector output. The name signals its focus: a Pro tier of Recraft's V4 Styles line that produces SVG rather than raster images. 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 only. Its context window is 65536 tokens, with a maximum output of 58982. The defining behaviour is style consistency: every request requires at least one style reference image, and the model generates a new image that reproduces that style. Output is vector, and the only supported output format is svg. Aspect ratio is selectable from a fixed set: 1:1, 2:1, 1:2, 3:2, 2:3, 4:3, 3:4, 5:4, 4:5, 16:9, 9:16, or auto. You can request between 1 and 6 images per call, and supply between 1 and 10 input references.

Those capabilities point to a fairly specific set of jobs. Because output is SVG, the results scale cleanly and can be edited as vectors, which suits logos, icons, illustrations, and other assets that need to hold up at different sizes or be recoloured and reshaped downstream. The style reference requirement makes it a good fit when you already have a visual language to match: brand guidelines, an existing icon set, or a house illustration style that new assets must follow. Supplying up to 10 references lets you describe that style more fully than a single example would, and generating up to 6 images per request helps when you want variations on the same brief. Text prompts drive the content, while the reference images anchor the look.

You call it through api14's OpenAI-compatible API using the model id recraft/recraft-v4-styles-pro-vector. The request takes text and image inputs and returns image output. Beyond the model id, the request supports image parameters rather than a general sampling parameter set: aspect_ratio (one of the twelve enum values listed above, including auto), output_format (svg), n (1 to 6), and input_references (1 to 10). Note that the supported_parameters list is empty, so the image parameters are the ones that apply here. At least one style reference image is required on every request, so plan for that in your payload rather than sending a text prompt alone.

The main constraint to weigh is the fixed output format. If you need raster images such as PNG or JPEG, or a model that works from text alone without a style reference, this model is not the match; its value is in vector output tied to a supplied style. If your pipeline can consume SVG and you have reference material to define the look, it fits that slot directly.