x
Generative artificial intelligence has taken another massive leap forward, moving beyond simple text-to-image prompts into dynamic motion and physical reality. Black Forest Labs, the pioneering German AI research organisation behind the celebrated FLUX series, has officially unveiled FLUX 3.
Marking a dramatic evolution from its predecessors, FLUX 3 transitions the company from still image synthesis into high-fidelity AI video generation. Crucially, the system extends far beyond digital screen content: the exact same AI engine is already being deployed to teach physical robots how to work on car manufacturing lines.
While previous iterations of the FLUX architecture dominated the AI artwork space, FLUX 3 introduces true multimodality. Rather than stringing together separate visual and audio engines, Black Forest Labs trained a single core network on images, video, and audio simultaneously.
This unified architecture enables FLUX 3 to generate high-definition video clips up to 20 seconds in length alongside perfectly synchronised sound. Whether it is spoken dialogue, ambient atmospheric sounds, or crisp acoustic effects, the audio is inherently tied to the visual action unfolding on screen.
Early blind preference evaluations demonstrate how competitive FLUX 3 is against top-tier industry rivals:
Alongside its video capabilities, FLUX 3 retains its legacy as a top-tier image generator, capable of producing diverse artistic styles and photorealistic detail across a broad range of prompts.
What makes FLUX 3 revolutionary is how its creators view video prediction. According to Black Forest Labs co-founder and CEO Robin Rombach, an AI model that only studies static images can only produce static images. By training a model to predict how pixels flow across a video sequence, the AI effectively learns the underlying physics of our world—comprehending concepts like momentum, contact, weight, and timing.
This physical understanding forms the foundation of FLUX-mimic, a specialised robotics model developed in partnership with Zurich-based mimic robotics.
FLUX-mimic connects the visual understanding of FLUX 3 to a lightweight "decoder" module. This decoder translates the model’s internal prediction of movement directly into robotic motor commands. Operating with an astonishing reflex time of roughly 101 milliseconds—comparable to human visual reaction speed—it allows robotic arms to react to fluid, real-world environments in real time.
Traditional industrial automation excels at rigid, highly repeatable tasks such as welding metal frames or spot-drilling chassis. However, conventional robots historically struggle with soft, deformable materials that require fine tactile adjustments.
German automotive giant Audi is currently testing FLUX-mimic on its production lines to solve this exact problem. The system is being used to manipulate flexible door seals—a delicate job requiring adaptive hand movements that previously required human precision.
By uniting generative AI visual predictions with physical robotics, FLUX-mimic allows machines to handle complex, soft-body manipulation tasks that were once considered un-automatable.
The launch of FLUX 3 represents an ambitious comeback for Black Forest Labs. Founded in August 2024 by key researchers who originally created Stable Diffusion at Stability AI, the lab quickly disrupted the industry with its initial FLUX 1 release, outperforming rivals like Midjourney and Stable Diffusion 3.
Although FLUX.2 launched in late 2025, competition intensified rapidly as alternative open-source tools gained traction. FLUX 3 re-establishes Black Forest Labs at the cutting edge of generative research, showcasing how generative media models can serve as fundamental engines for spatial intelligence and robotics.
FLUX 3 is currently rolling out in early access. The Video and Action capabilities are available via restricted APIs and select enterprise partners, including mimic robotics and Audi. Expanded image generation capabilities are set to follow in the coming weeks.
For open-source enthusiasts and local deployment, Black Forest Labs plans to release the open-weight FLUX 3 Dev version later in 2026.
To read the original report and explore additional context regarding this development, visit the full news coverage at Decrypt:
👉 Black Forest Labs Unveils FLUX 3 AI: Ditches Stills for Video—And Robot Hands
Disclaimer: This article is provided for informational purposes only, mistakes may be made, and it's not offered or intended to be used as legal, tax, investment, financial, or any other advice.
