Recraft: Recraft V4 Pro

Recraft: Recraft V4 Pro is an image generation model from Recraft, listed on api14 under the model id recraft/recraft-v4-pro. It sits in the image category and is aimed at developers who need generated images rather than text completions. Recraft is the provider behind the model, and the version name indicates this is the Pro tier of the V4 line.

The model accepts text and image inputs and returns images as output. Its context window is 65,536 tokens, with a maximum output of 58,982 tokens. Generated images come out at roughly 2K resolution, and the model supports several aspect ratios: 1:1, 4:3, 3:4, 16:9, 9:16 and auto. You can request between 1 and 6 images in a single call, and you can supply up to one input reference image alongside your text prompt. The reference input is useful when you want the output to follow the style, subject or composition of an existing image rather than relying on text alone.

Those capabilities point to a few practical uses. Product and marketing teams can generate square or widescreen visuals for campaigns, with the aspect ratio chosen to match the placement. Designers can produce multiple variations of a concept in one request by setting n above 1, then pick the strongest result. Because image input is supported, workflows that start from a sketch, a photo or a brand asset can use it as a reference and generate a new image in a similar direction. The 2K output resolution suits web and print layouts where a low resolution draft would not hold up.

You call the model through api14's OpenAI-compatible API using the model id recraft/recraft-v4-pro. Because the interface follows the OpenAI format, existing clients and SDKs work with minimal changes. The model exposes no general supported parameters, so the controls live in the image parameters instead: aspect_ratio, which takes one of the enum values listed above, n, which ranges from 1 to 6, and input_references, which accepts 0 or 1 reference image. Set those alongside your prompt in the request.

One thing to keep in mind is that this is an image-only model. It does not return text, so tasks like summarisation, classification or chat belong with a text model instead. If you need more than one reference image, or aspect ratios outside the six listed values, this model will not cover that case. Within its scope, though, it handles text-to-image and reference-guided generation at 2K across the supported ratios.