Skip to main content

Black Forest Labs' Flux3: AI That Generates 20-Second Videos with Sound

German AI startup Black Forest Labs has officially launched Flux3, a multimodal foundation model that pushes the boundaries of generative AI. Unlike many models that handle only images or videos, Flux3 is built on the company's Self-Flow architecture, integrating dedicated codecs for images, video, audio, and motion. This allows it to understand and generate content across physical and digital environments in a unified way.

Native Audio and Video Generation

One of Flux3's standout features is its ability to generate synchronized audio and video clips up to 20 seconds long in a single pass. It supports a range of functions, including text-to-video, image-to-video, video-to-video, keyframe-based transitions, and even multilingual dialogue. In early testing at 720p resolution with 10-second clips, Flux3 achieved a 93% win rate against Luma Ray3.2, crushed Runway Gen-4.5 with a 77% win rate, and held a slight edge over top models like Seedance 2.0 and Gemini Omni Flash.

Image

Stepping into the Real World

Black Forest Labs isn't stopping at digital content. The company partnered with Mimic Robotics to develop Flux-mimic, a video action model for robotics. It's already being tested in an Audi factory, where it handles production tasks. This move extends AI capabilities from the purely digital realm into physical manufacturing scenarios.

Phased Release Strategy

BFL is rolling out Flux3 in stages. Flux3 Video is available now, while Flux3 Image and the open-source weight "Flux3 Dev" are slated for release in the near future. This approach allows developers and researchers to start experimenting with the video capabilities immediately.

Key Points

  • Flux3 generates up to 20-second videos with native, synchronized audio.
  • It outperforms Luma Ray3.2 (93% win rate) and Runway Gen-4.5 (77% win rate) in early tests.
  • The model uses Self-Flow architecture with dedicated codecs for multiple modalities.
  • A robotics variant, Flux-mimic, is being tested in an Audi factory.
  • Flux3 Video is available now; image and open-source versions are coming soon.