FLUX is a family of generative AI models made by Black Forest Labs. The models create and edit images from a text prompt. The newest generation, FLUX 3, is one multimodal model that also makes video clips with sound and predicts robot actions.
The FLUX.1 models work with flow matching in a latent space: the model learns to turn random noise, step by step, into a compressed representation of an image, and an autoencoder then decodes it into pixels. Besides text-to-image, there are versions for inpainting and outpainting (filling or extending part of a picture), for following the edges or depth of a reference image, for image variations, and for editing an image from an instruction (FLUX.1 Kontext). Open-weight models can be downloaded from Hugging Face, then run, fine-tuned and customized on a company's own servers. The Pro models are available only through the API.
FLUX is offered through a paid API, billed per image or per megapixel, and as open weights: FLUX.1 [schnell] under the Apache 2.0 license, the [dev] models under a non-commercial license with a paid license for commercial use.