What it does

FLUX is the model lineup from Black Forest Labs. The current generation, FLUX 3, is presented as one multimodal model covering image, video, audio and action prediction, with separate pages for FLUX 3 Image, Image Klein, Video and Action, plus a set of FLUX Tools. On the video side, the site describes clips of up to 20 seconds in a single generation, started from text, an image or keyframes, with several shots per take and optional audio such as multilingual speech, effects and ambience. Image output is described as stylistically varied, with accurate text rendering and good handling of complex prompts. FLUX 3 Action is an open weights 7B model aimed at robotics: it takes visual observations and text instructions and predicts the outcome visually and as robot control actions. There are three ways in: try the models in a browser playground, integrate through the FLUX API for production workloads, or get a licence for the open weights and run them on your own infrastructure with room to fine-tune and customise. Enterprise licensing for the API and open weights is offered separately, and the company lists SOC 2 and ISO 27001 certification.

Capabilities

Image, video, audio and action prediction in the FLUX 3 family
Video clips up to 20 seconds with optional generated audio
Image generation with accurate text rendering across many styles
Hosted API for production workloads
Open weights to run, fine-tune and deploy on your own infrastructure
Browser playground for trying models without code

When to use it

You want to add image or video generation to a product through an API
You need model weights you can host and customise yourself, subject to a licence
You want to try the models in a browser before committing to an integration
You work on robotics and want a model that predicts actions from visual input