Paper 2512.16093
TurboDiffusion: Accelerating Video Diffusion Models by 100-200 Times
- Published
- Dec 2025
- Research lab
- Independent
- Citations
- 28
- GitHub
- 3.6K stars
01 In brief
Summary
TurboDiffusion is a video generation acceleration framework that achieves 100–200× end-to-end speedup while maintaining video quality.
It combines four main techniques: low-bit SageAttention for attention acceleration, Sparse-Linear Attention (SLA) for sparse attention, rCM for step distillation, and W8A8 quantization for linear layers.
Training involves finetuning with SLA and distilling with rCM, then merging updates.
Inference uses SageSLA, reduces sampling steps to 3–4, quantizes parameters and activations to INT8 with block-wise granularity, and reimplements operations like LayerNorm in Triton/CUDA.
Evaluated on Wan2.2-I2V-A14B-720P, Wan2.1-T2V-1.3B-480P, Wan2.1-T2V-14B-720P, and Wan2.1-T2V-14B-480P on a single RTX 5090, TurboDiffusion reduces latency from 4549s to 38s, 184s to 1.9s, 4767s to 24s, and 1676s to 9.9s respectively, achieving speedups of 120×, 97×, 199×, and 170×.
It outperforms FastVideo in both efficiency and quality.
Future work includes extending to autoregressive video diffusion.
02 From the paper
Abstract
We introduce TurboDiffusion, a video generation acceleration framework that can speed up end-to-end diffusion generation by 100-200x while maintaining video quality. TurboDiffusion mainly relies on several components for acceleration: (1) Attention acceleration: TurboDiffusion uses low-bit SageAttention and trainable Sparse-Linear Attention (SLA) to speed up attention computation. (2) Step distillation: TurboDiffusion adopts rCM for efficient step distillation. (3) W8A8 quantization: TurboDiffusion quantizes model parameters and activations to 8 bits to accelerate linear layers and compress the model. In addition, TurboDiffusion incorporates several other engineering optimizations. We conduct experiments on the Wan2.2-I2V-14B-720P, Wan2.1-T2V-1.3B-480P, Wan2.1-T2V-14B-720P, and Wan2.1-T2V-14B-480P models. Experimental results show that TurboDiffusion achieves 100-200x speedup for video generation even on a single RTX 5090 GPU, while maintaining comparable video quality. The GitHub repository, which includes model checkpoints and easy-to-use code, is available at https://github.com/thu-ml/TurboDiffusion.