arXiv Project Page

TempoMaster is a video diffusion model built on Wan-Video, which is capable of generating videos at various frame rates.

The model first generates a low-frame-rate video as a global blueprint. It then uses the existing frames as temporal anchors to infer and insert additional frames in between, progressively upsampling the video to higher frame rates.

This approach effectively structures long-term temporal dynamics and mitigates the issue of visual drifting caused by error accumulation.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Model tree for Scottttttyy/TempoMaster

Finetuned
(42)
this model