FLUX3: AI video made in Europe

Europe is taking a significant step in the AI race. Black Forest Labs, the German company founded by the creators of Stable Diffusion, is introducing FLUX3: the first Europe-developed multimodal model that combines video, audio, and action prediction within a single architecture. The model generates clips up to 20 seconds long with native audio, supports multiple input types (text, images, keyframes), and in early tests outperforms competitors such as Runway and Luma.

One model, four modes

FLUX3 is no ordinary video generator. It's a unified model that learns simultaneously from images, video, audio, and physical actions, built on Black Forest Labs' Self-Flow architecture. Where other models use separate pipelines for sound and image, FLUX3 trains everything at once. The result? Natural synchronization between visual events and sound — a falling glass sounds like glass, a slamming door sounds like a slamming door. The model supports text-to-video, image-to-video, video-to-video, and keyframe-to-video.

FLUX3 is more than a technological breakthrough; it's a strategic move for Europe. As the first EU-developed multimodal video generator, it aligns with Europe's ambition to reduce dependence on American and Asian AI. The EU AI Act and the Open Source Strategy emphasize the importance of local control over data, infrastructure, and operational chains. FLUX3 meets this: the model is available via API, open weights (later in 2026), and a playground.

From content creation to robotics

FLUX3 isn't intended for video creators alone. The model is already being deployed in the automotive industry, where FLUX-mimic — a variant for action prediction — works together with Mimic Robotics and Audi. Here it solves complex tasks, such as manipulating soft materials (seals, cables), which have traditionally been difficult to automate. Its reaction time of roughly 100 ms approaches human visual response times, which is essential for real-time applications. For the media industry, FLUX3 offers native audio integration, multilingual dialogue, and stylistic diversity (from camcorder footage to animation), making it ideal for branded content, product videos, and visual effects (VFX).

In early evaluations, FLUX3 performs more strongly than many competitors. In the coming months, FLUX3 Image, FLUX3 Action, and an open-weight version (FLUX3 Dev) will be released, further democratizing access.