ForcingBooster: Adaptive Patch-wise Denoising for Fast Autoregressive Video Diffusion
ForcingBooster cuts autoregressive video diffusion steps about 50% by stopping easy patches early.
ForcingBooster dynamically allocates denoising steps across spatial patches in autoregressive video diffusion, letting easy regions stop early while harder regions receive more refinement. It adds a lightweight routing module that works with most existing few-step autoregressive diffusion models and needs only modest training. Experiments cut the total denoising steps applied across patches by about 50% while keeping visual quality comparable to a four-step baseline.
- Assigns denoising steps per patch instead of a uniform schedule.
- Easy patches stop early; harder patches get extra refinement.
- Needs only a lightweight routing module and modest training.
- Cuts total patch denoising steps by about 50% versus a four-step baseline.
- Compatible with most few-step autoregressive video diffusion models.
From the original article
1 min read at huggingface.co
Autoregressive video diffusion models have become a leading approach to real-time interactive video generation and video world modeling. However, existing methods typically allocate computation uniformly across space when generating the next video chunk, updating all patches with the same denoising schedule and number of function evaluations.
By Zekai Gu, Peng Wang, Jiahao Lu · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.