Black Forest Labs has released FLUX 3, its first video-generation model, and the same system is already being used to train robots on an Audi assembly line. The move marks a shift from the company's previous focus on still-image generation into the more complex domain of moving pictures — and into industrial automation.
From images to video
FLUX 3 is the latest in the FLUX family of generative AI models. Earlier versions, such as FLUX.1, were designed to create and edit static images. The new model extends that capability to video, producing short clips from text prompts or from existing footage. Black Forest Labs has not disclosed the exact resolution or frame rate of the output, but the company describes it as a general-purpose video generation tool.
The model was trained on a large dataset of videos, though the company has not specified the sources or size of that dataset. Black Forest Labs says FLUX 3 can handle tasks like video-to-video translation, where a user provides a source clip and a text description of the desired change, and the model alters the footage accordingly.
Robots learning on the line
In a parallel application, the same FLUX 3 system is being used to train robots on an Audi assembly line. The robots learn by watching video generated by the model, which simulates various assembly scenarios. This approach allows the robots to practice tasks without needing physical prototypes or risking damage to real equipment.
Audi has not provided details on which specific assembly processes are being trained, how many robots are involved, or when the system might move from pilot to production. The automaker said only that the collaboration with Black Forest Labs is part of a broader effort to make its factories more flexible and efficient.
Using generative video for robot training is a relatively new technique. Instead of programming each movement manually, engineers feed the model a description of the task — say, installing a door panel — and the model generates a video of the correct sequence. The robot then mimics that sequence, adjusting for variations in the real environment.
What FLUX 3 can do
Black Forest Labs has positioned FLUX 3 as a tool for both creative professionals and industrial users. On the creative side, the model can generate short video clips from text prompts, similar to products from companies like OpenAI and Runway. On the industrial side, the model's ability to produce consistent, repeatable video sequences makes it suitable for training robots.
The company has not announced pricing or a public API for FLUX 3. It is currently available only to select partners, including Audi. Black Forest Labs says it plans to release more details about the model's capabilities and availability later this year.
The launch comes as competition in the generative video space heats up. OpenAI's Sora, released earlier this year, can generate minute-long videos from text. Runway's Gen-3 Alpha is also widely used. Black Forest Labs is betting that its focus on both creative and industrial use cases will give it an edge.
Neither Black Forest Labs nor Audi has said how long the robot training pilot will last or what metrics will determine success. The companies are expected to share more information at an industry conference in the coming months.
