ByteDance's Seedance 2.5 claims 30-second AI videos — independent tests still needed

ByteDance's new AI video model, Seedance 2.5, has officially arrived. According to a report from International Business Times Singapore, ByteDance claims the model can generate AI video clips of up to 30 seconds — but independent tests have yet to verify those claims.

Seedance 2.5 is the successor to Seedance 2.0, which led the image-to-video leaderboard for models with audio on Artificial Analysis. The headline change is that video and audio are generated in a single pass, producing clips up to 30 seconds long that can be extended multiple times.

For scene construction, users can upload up to 30 images, 10 video clips, and 10 audio files as reference material to create shots with multiple characters and camera angles. ByteDance also says textures, lighting, and skin detail have improved in the new version.

Seedance 2.5 is live on Jimeng AI and Doubao Pro, with API access through BytePlus ModelArk coming later. As a showcase, ByteDance produced an entire short film with the model, demonstrating how ad and visual-content teams could build full productions in one pipeline instead of stitching together individual clips.

Duration is a key variable in this race. The claimed 30-second ceiling is well above the single-generation length of some mainstream rivals — Google's Gemini Omni Flash tops out at about 10 seconds — making the model more attractive for one-shot ads, short videos, and trailers.

The reporting, however, remains cautious: these capabilities are vendor claims that have not been independently verified. There is often a gap between a video model's marketed specs and its actual output, and whether 30-second clips hold up on consistency, motion coherence, and audio sync is a question only third-party evaluation can answer.

The next things to watch are the API's opening date, independent benchmark results, and how Seedance 2.5 compares with rivals like Google in real workflows. For content teams, if the one-pass, audio-ready promise holds, the video production pipeline will shrink dramatically.