Multimodal direction
Work from text, first or last frames and supported image references.
Model guide
Seedance 2.5 for reference-rich cinematic motion.
A flexible route for text, frames and image references when the shot needs visual direction and native sound.
What it gives you
Use it when the direction contains more than a prompt.
Seedance 2.5 combines longer duration control with first and last frames, multiple references and native audio. Soho exposes only the configurations it has mapped and priced.
Longer clips
Choose from the active duration range instead of a single fixed length.
Native sound
Generate with the model's supported audio behavior when enabled.
Model details
Seedance 2.5
Text, first/last frame, and reference-guided clips from 4–30 seconds with selectable resolution and native audio.
- Inputs
- Text · First frame · Last frame · Reference images
- Duration
- 4–30s
- Resolution
- 480p · 720p · 1080p
- Aspect ratios
- adaptive · 16:9 · 9:16 · 1:1 · 4:3 · 3:4 · 21:9
- Reference images
- Up to 30
- Native audio
- Yes
- Availability
- Available
Generated examples
Generated examples
Real Soho Frame examples identified by the model used.
Kingdom Rising
Seedance 2.5
Liberty Colossus
Seedance 2.5
The Last Crown
Seedance 2.5
A clear workflow
A clear workflow
- 01
Set the input
Choose text, frames or references from the inputs the model actually accepts.
- 02
Shape the output
Select duration, resolution, ratio and audio where those controls are available.
- 03
Review in the project
The result returns beside its source with the generation settings attached.
Questions
Questions
How many reference images does Seedance 2.5 accept?
The active Soho catalog currently documents up to 30 reference images. Frame modes and reference modes have different controls.
Does it support audio?
Yes. Native audio is supported in the mapped Soho route.
Is every provider feature exposed?
No. Source-video, source-audio and automatic-duration modes remain gated until their validation and pricing contracts are complete.
Try it on your next shot.
Open Video with the model selected and keep the result in its project.
Create with this model