ByteDance's Seedance 2.5 doubles single-pass generation from 15 to 30 seconds and keeps extending from there, so a setup-to-resolution story lands in one take instead of a stitched timeline. It accepts up to 30 images, 10 video clips and 10 audio clips as references in a single pass, and edits down to a timestamp.
Capability snapshot
30s
single-pass clip
30+10+10
image, video, audio refs
timestamp
level editing
Practical value mapped to real production workflows.
Benefit
01
A whole story in one pass
Thirty seconds is long enough for setup, development, a turn and a resolution across several connected shots, rather than one extended moment. Multi-round extension continues from an existing output while holding characters, environment and pacing, so multi-minute pieces stop being a splicing job.
Benefit
02
References that carry intent
Up to 30 images, 10 video clips and 10 audio clips go in together, covering clay render, motion and creative references. More material and more kinds of it give the model a firmer grip on what you meant across multiple subjects, scenes and shot changes.
Benefit
03
Edits that land where you point
Timestamp-level control targets a specific stretch of audio or video instead of regenerating the clip, and green screen, camera perspective and reference-based editing cover the demands of film and advertising work.
A simple three-step path from source material to a reusable result.
Step 1
Gather your references
Bring the images, clips and audio that define your subjects, motion and sound, up to 30 / 10 / 10 in one pass.
Step 2
Write the shot sequence
Describe the shots in order and how they connect. Seedance 2.5 organizes them into one continuous take with synced audio.
Step 3
Extend or edit
Append the next 30 seconds with multi-round extension, or fix a stretch in place with timestamp-level editing.
Up to 30 seconds in one pass, which is double Seedance 2.0's 15. Multi-round extension appends further clips to an existing output while keeping characters, scene and sound consistent, so multi-minute results are reachable without manual splicing.
Images, video and audio, together in the same request — up to 30 images, 10 video clips and 10 audio clips. Reference types include clay render, motion and creative references.
It keeps 2.0's joint audio-video architecture and builds on it: single-pass length goes from 15 to 30 seconds, reference input widens to images plus video plus audio, shot transitions and scene changes improve for long-form continuity, and editing gains timestamp-level control.
Next step
Open the tool with a real image, compare the results, and keep the version that works.