Recraft: Recraft V4

Recraft V4 is an image generation model from Recraft, listed on api14 under the model id recraft/recraft-v4. It sits in the image category, so its purpose is producing images rather than text, and it comes from a provider that specialises in visual generation tooling. If you already route other models through api14, this one slots into the same catalogue alongside your text and multimodal models.

The model accepts text and image inputs and returns image output at roughly 1K resolution. Its context window is 65,536 tokens, with a maximum output of 58,982 tokens. The image parameters give you control over the shape and count of what comes back: aspect_ratio accepts 1:1, 4:3, 3:4, 16:9, 9:16 or auto, n sets how many images are generated in a single request between 1 and 6, and input_references takes between 0 and 1 reference images. Recraft describes the model as having stronger compositional judgment, which matters when a prompt describes several objects and their arrangement rather than a single subject.

Those capabilities point to a fairly specific set of jobs. Text to image generation covers the usual ground: concept art, blog and article illustrations, social media assets, and placeholder or production imagery for product pages. Because the model accepts an image input alongside text, you can pass a single reference image and steer the result with a prompt, which suits style matching, variations on an existing asset, or adapting a visual to a different aspect ratio. The aspect ratio options map directly onto common delivery formats, so a 16:9 hero image and a 9:16 story asset can come from the same model without post processing crops. Generating up to six images per call is useful when you want a small set of options to choose from before committing to one.

Calling it through api14 works the same way as any other model on the platform, using the OpenAI-compatible API with the model id recraft/recraft-v4. Note that the model exposes no general supported parameters, so the tuning happens through the image parameters described above rather than through the usual sampling controls. Set aspect_ratio to one of the listed values or leave it on auto, set n between 1 and 6, and supply at most one input reference when you want the output grounded in an existing image. The response contains image output, so plan your client around handling image data rather than text completions.

The main constraint to keep in mind is that this is a single purpose image model. It does not return text, and with no supported parameters there is little room to adjust behaviour beyond aspect ratio, image count and the single reference input. If your task is text generation, reasoning, or anything that needs a text response, a different model in the catalogue will fit better. If you need more than one reference image, or finer control over generation settings, check whether another image model on api14 exposes those options. For straightforward text to image work and single reference variations at around 1K resolution, Recraft V4 covers the common cases.