Step-Video TI2V
Step-Video TI2V accepts a starting image and a text instruction. StepFun published its weights and inference code on 17 March 2025.
Control motion separately from the starting image
The official examples vary motion intensity and show camera instructions such as tracking, orbiting and zooming. These controls guide how the initial image develops over time.StepFun Step-Video-TI2V official repository
Check identity across the clip
Practical guidance. The starting image anchors appearance, but moving or turning objects can still change. Inspect the final frame and any large pose change, not only the initial view. Keep movement instructions consistent with the source composition.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Launch date | |
| Access | Downloadable |
| Model identifiers | stepvideo-ti2vStep-Video-TI2V |
| Parameters and length | The official code repository specifies 30B parameters and up to 102 frames.StepFun Step-Video-TI2V official repository |
| License | MIT.StepFun stepvideo-ti2v model card |
| Motion control | The pipeline exposes a motion_score parameter and prompt-based camera directions.StepFun Step-Video-TI2V official repository |
Related models
- Step-Video T2V
StepFun · Step-Video
Step-Video T2V is a 30B text-to-video model that can generate up to 204 frames.
Launch date Downloadable - Step-Video T2V Turbo
StepFun · Step-Video
Step-Video T2V Turbo is the inference-step-distilled version of Step-Video T2V.
Launch date Downloadable