4DAnyone Multi-View Video
Use one portrait video to synthesize synchronized camera views of the same person. Download the videos, camera rig and metadata for a separate reconstruction workflow.
The current multi-view demo requires a long GPU allocation and generally needs PRO quota. Existing examples can be viewed for free. The output is a set of videos and camera data; creating a 4D reconstruction is a separate step.
Try the tool
What you need
- Video
Generate New Views of a Performance
Loading the demo…
How to use it
1. Prepare your inputs
Open the demo and add the inputs listed above. You can start with an example inside the demo before uploading your own files.
2. Set up the result
Use a 9:16 clip with one fully visible person and a stationary camera, with at least 121 frames. Select 1, 2, 4 or 6 views and adjust the camera ring. More views require more GPU time.
3. Generate and review
Run the selected task and wait for the queue. Review the result, adjust one setting at a time, and save the output using the controls in the demo.
Common questions
Is the demo free to use?
The current multi-view demo requires a long GPU allocation and generally needs PRO quota. Existing examples can be viewed for free. The output is a set of videos and camera data; creating a 4D reconstruction is a separate step.
Where are my uploads processed?
The selected Hugging Face Space receives your inputs and runs the model. Its maintainer controls processing and retention. The surrounding guide is translated; the demo uses the languages provided by its maintainer.
What if the demo does not load?
Reload the workspace or choose a backup server if available. If your GPU quota is used up, wait for it to reset or try a related tool.