ByteDance's New Video Model Seedance 2.5 Generates 30-Second Clips with Integrated Audio

ByteDance has launched Seedance 2.5, a new version of its AI video model that generates both video and audio in a single pass, creating clips up to 30 seconds that can be extended multiple times. This outperforms competitors like Google's Gemini Omni Flash, which currently manages only about 10 seconds at a stretch. Users can upload up to 30 images, 10 video clips, and 10 audio files as reference material to design scenes with multiple characters and camera angles. The model features improved textures, lighting, and skin detail. Its predecessor, Seedance 2.0, leads the image-to-video leaderboard for models with audio, according to Artificial Analysis. ByteDance showcased the model's capabilities in a short film fully produced with Seedance 2.5, demonstrating a seamless production workflow for advertising and visual content. Seedance 2.5 is available on Jimeng AI and Doubao Pro, with API access via BytePlus ModelArk to follow. Previously, 'District 9' director Neill Blomkamp showed what the earlier version could do with his 13-minute fully AI-generated short film 'Nightborne'.