ByteDance's Seedance 2.5 generates 30-second video clips with built-in audio

ByteDance has released Seedance 2.5, a video generation model that produces synchronized video and audio together rather than treating them as separate outputs to be combined later. The clips it generates run up to 30 seconds - three times the length of what Google's Gemini Flash Omni currently offers - which puts it toward the higher end of what commercially available models can produce in a single generation.

One of the more notable aspects of Seedance 2.5 is its multimodal input support. Users can provide dozens of reference files at once, spanning still images, existing video footage, and audio recordings. That kind of flexible reference system allows for more directed outputs, giving creators and production teams a way to anchor generated content to specific visual or sonic materials rather than working purely from text prompts.

The integrated audio generation is a meaningful step. Most video AI tools either omit audio entirely or require a separate model and a manual sync step afterward. Generating both in a single pass reduces friction in the pipeline and keeps audio and video naturally timed to each other from the start, which can be difficult to achieve when the two are produced independently.

ByteDance has framed part of the appeal around advertising production, where teams often need to assemble short clips into a finished piece. If a single model can output a polished 30-second segment with usable audio, that collapses several steps in the typical workflow. Whether Seedance 2.5 delivers on that in practice - particularly for brand-specific or heavily art-directed work - will depend on how much control the reference inputs actually provide and how consistent the outputs are across a production run.