Extending a Clip on SceneWish: Upload Your Own or Continue a Finished Task
ChaoziSceneWish can continue a video clip with AI, but the first decision on the Extend Video page is not the prompt. It is the source. Extension models fall into two groups, and the model decides which source you can use.
Models that take an uploaded clip
- Seedance 2.5 (ByteDance): 1 to 10 videos of 2 to 30 s each, up to 30 s in total; images and audio optional. Output 4 to 30 s, up to 1080p.
- Wan 2.7 (Alibaba): one video of 2 to 10 s. Output 2 to 15 s, 720p or 1080p.
- SkyReels V4 Fast (Skywork): one video of 1 to 15 s. Output 3 to 15 s, up to 1080p.
- FLUX 3 Video (Black Forest Labs): one video of 1 to 30 s. Output 5 to 20 s, 720p or 1080p.
Models that only continue their own finished task
- Veo 3.1 Fast (Google): continues a finished 8-second Veo 3.1 Fast video into a 15-second result, up to 4K.
- PixVerse V6: continues a finished PixVerse V6 task, 1 to 15 s.
- Gemini Omni 1.1 Flash (Google): continues a finished Gemini Omni task, and the model picks the length.
Then describe what happens next
Write the next action, and keep the same subject, setting and style unless you want them to change. Pick the length the model offers and check the credit quote before generating.
A note on the examples
The demonstrations on the page, such as a solo dancer joined by a group, are labelled as external examples from the model makers, not SceneWish outputs.
Written by the SceneWish team.