Black Forest Labs makes FLUX 3 Video generally available and claims it beats Seedance 2.0

Black Forest Labs has made FLUX 3 Video generally available, bringing its latest video generation model to a wider audience after what appears to have been a limited early access period. The model produces Full HD clips with a maximum duration of 20 seconds - a meaningful step up in length and resolution compared to many competing offerings that still cap out at shorter runtimes or lower resolutions.

One of the more technically notable additions is native audio generation paired with lip-synced dialogue. Rather than requiring a separate audio layer to be composited onto generated footage, FLUX 3 Video handles both simultaneously, and the lip-sync capability works across more than 14 languages. That kind of multilingual support at the generation stage - rather than as a post-processing dubbing step - reduces friction for creators working across different regional markets.

The model also introduces in-scene typography rendering, allowing readable text to appear naturally within generated video frames. Legible text in AI-generated images and video has historically been a weak point for diffusion-based models, so its inclusion as a feature worth highlighting suggests BFL has made targeted improvements in that area.

On the competitive positioning side, Black Forest Labs is citing its own Elo-based ranking system to place FLUX 3 Video above Google's Gemini Omni Flash and ByteDance's Seedance 2.0. Elo rankings derived from user preference votes can be a useful signal, but it is worth noting that BFL is both running and reporting these benchmarks, which warrants some caution in taking the comparisons at face value. Independent third-party evaluations will ultimately offer a clearer picture. Black Forest Labs, the team behind the widely used FLUX image generation models, has been working to expand into video, and this release marks a concrete step in that direction.