ant-research/4DAnyone
The repo now defaults to a Turbo model that cuts denoising time by 5.58x.
4DAnyone takes a casual single-camera video of a person and generates multi-view videos for 4DGS reconstruction. The README says Turbo and Base produce comparable generation quality, with Turbo enabled unless users turn it off. It supports configurable view counts, pitch layers, and yaw coverage, including dense 24-view and 48-view layouts. Custom inputs should be portrait video at 720p or higher, with at least 121 frames and only mild camera motion. GitHub · AI repos' note.
4DAnyone takes a casual single-camera video of a person and generates multi-view videos for 4DGS reconstruction. The README says Turbo and Base produce comparable generation quality, with Turbo enabled unless users turn it off. It supports configurable view counts, pitch layers, and yaw coverage, including dense 24-view and 48-view layouts. Custom inputs should be portrait video at 720p or higher, with at least 121 frames and only mild camera motion. GitHub · AI repos' note.
score 5