ByteDance Unveils Seedance 2.5 With 30-Second Generation And Multi-Modal Input Support

ByteDance releases Seedance 2.5, doubling generation time to 30s and enabling timestamp editing. The model supports complex multi-modal inputs and is rolling out via Dream AI and Volcano Ark.

Woofun AI reports that ByteDance’s Seed team has launched Seedance 2.5, a video generation model that extends single-generation duration from 15 to 30 seconds. The system maintains joint input architecture for text, images, video, and audio, allowing users to extend clips while preserving consistency in characters, scenes, and narrative pace.

The model accepts up to 30 images, 10 video clips, and 10 audio clips simultaneously. It introduces timestamp-based control, enabling precise adjustments to actions, camera shots, or audio without regenerating entire segments. Seedance 2.5 is currently rolling out on Dream AI and Bean Professional Editions, with API integration into Volcano Ark imminent.