This is a very solid and promising research that scales consistency models to 10B+ video diffusion models. The combination of sCM and Variational Score Distillation is a very promising direction for few-step generation!
Kaiwen Zheng @zkwthu
Oct 10
馃殌Try out rCM鈥攖he most advanced diffusion distillation! 鉁匜irst to scale up sCM/MeanFlow to 10B+ video models 鉁匫pen-sourced FlashAttention-2 JVP kernel & FSDP/CP support 鉁匟igh quality & diversity videos in 2~4 steps Paper: arxiv.org/abs/2510.08431 Code: github.com/NVlabs/rcm

Oct 10, 2025 路 6:51 PM UTC

2
12
123
Replying to @clu_cheng
Huge thanks to @clu_cheng for the fantastic sCM work last year鈥攕uch a promising direction for speeding up diffusion model inference!
1
Replying to @clu_cheng
This innovative approach could significantly advance the field of video generation. Excited to see how these methods evolve and their potential applications emerging!
Load more