Seedance 2.5
ByteDance's 30-second audio-and-video model with timeline editing of individual sections.
Quick verdict
- Best for
- Creators who want to direct rather than re-roll
- Pricing
- Via Jimeng AI and Doubao Pro subscriptions; per-generation pricing on partner platforms
- Not ideal if
- Primary access is through ByteDance's Chinese apps
What is Seedance 2.5?
Seedance 2.5, released by ByteDance on July 31, 2026, generates 30 seconds of synchronised audio and video in one pass, twice the length of Seedance 2.0, at native 4K. Creators can feed it up to 30 images, 10 videos, and 10 audio clips, then use timestamps to direct or revise individual sections, which moves it from clip generator toward an editable production system. It is rolling out through Jimeng AI and Doubao Pro and is available through partner platforms.
Key features
- 30-second synchronised audio and video
- Native 4K
- Up to 30 images, 10 videos, and 10 audio clips as input
- Timestamp-based direction and revision of sections
- Available through partner platforms
Pros
- Section-level revision is closer to editing than regenerating
- Audio and video in one pass
- Rich multimodal input for consistency
Cons
- Primary access is through ByteDance's Chinese apps
- Pricing varies by platform
- Same content-policy and data questions as other Chinese-hosted models
Best for
Read more
Related comparisons
Alternatives to Seedance 2.5
Kling 4.0
Kuaishou's next video model: 30-second 4K clips steered by up to 10 keyframes, Flash out first.
MiniMax H3
A general multimodal video model: text, image, video, and audio in, 15-second 2K clips with stereo out.
Wan3.0
Alibaba's 30-second video model that also generates from documents, slides, and web pages.