# TurboDiffusion(Nvidia GPU)

TurboDiffusion(Nvidia GPU) is a self-hosted ai app. TurboDiffusion is an open-source video-generation acceleration framework that speeds up end-to-end text-to-video diffusion by 100-200x (benchmarked on a single NVIDIA RTX 5090) while maintaining video quality. It is open source and self-hosted, installable on ZimaOS.

Canonical page: https://ourlocalhost.com/app/3833/turbodiffusion-nvidia-gpu

## Description

TurboDiffusion is an open-source video-generation acceleration framework that speeds up end-to-end text-to-video diffusion by 100-200x (benchmarked on a single NVIDIA RTX 5090) while maintaining video quality. It combines SageAttention and Sparse-Linear Attention for faster attention computation with rCM for timestep distillation, and it ships optimized, quantized and non-quantized checkpoints tuned for different GPU configurations so Wan-family video models turn text prompts into high-quality clips in a fraction of the usual time. It is aimed at self-hosters with a CUDA-capable NVIDIA card that has large video memory (VRAM) and who want fast local text-to-video generation without a multi-GPU rig, and it is worth using whenever you need dramatically shorter render times. Note that the bundled models are currently trained only on long English prompts.

## Platform availability

Available on 1 of 10 platforms: ZimaOS.

## Health score

59/100 (Fair), based on 4 of 4 factors.
- Maintained: 82 (last commit 17 days ago)
- Popular: 72 (3,859 GitHub stars)
- Easy to install: 35 (1 install platform, install notes, Docker image)
- Light to run: 10 (40 GB minimum RAM)

## Links

- GitHub: https://github.com/thu-ml/TurboDiffusion
- ZimaOS: https://github.com/IceWhaleTech/CasaOS-AppStore/tree/main/Apps/TurboDiffusion_Nvidia
- Official website: https://arxiv.org/pdf/2512.16093

## Common questions

### How much speedup does TurboDiffusion claim for diffusion video generation?

It claims 100 to 200 times faster end-to-end diffusion generation on a single RTX 5090.

### Which acceleration technique is called out as primary?

TurboDiffusion primarily uses SageAttention.

---
Source: Our Local Host (https://ourlocalhost.com/app/3833/turbodiffusion-nvidia-gpu), the self-hosted and open-source app directory.
