FlashVSR
FlashVSR is a free, self-hosted framework for streaming video super-resolution that uses a one-step diffusion process.
What FlashVSR is
FlashVSR is a free, self-hosted framework for streaming video super-resolution that uses a one-step diffusion process. Its design combines a three-stage distillation pipeline, locality-constrained sparse attention, and a tiny conditional decoder. The project is primarily optimized for 4× enhancement and recommends that setting for improved results and stability. It reports about 17 FPS for 768 × 1408 video on one A100 GPU. The repository offers v1 and v1.1 weights through Hugging Face and recommends v1.1; listed modes include v1, v1.1, full, tiny, and tiny long video. Batch processing and MP4, MOV, AVI, and MKV are listed. Running it requires a local Python environment using Python 3.11.13 and the Block-Sparse Attention backend. That backend is reported to work correctly on A100 and A800 GPUs; H200 has limited acceleration, while compatibility on other listed GPUs is unknown. The project warns that implementations without its sparse attention module may lose noticeable quality, particularly at higher resolutions. The repository identifies the license as Apache-2.0.
Good
- Open-source code and model weights are free.
- Supports batch processing and four listed video formats.
- Project recommends v1.1 model weights.
- Apache-2.0 license.
Watch out
- Requires a local Python environment and attention backend.
- GPU compatibility varies; other listed GPUs are unknown.
- H200 runs with limited acceleration.
Who it suits
FlashVSR is for users prepared to run a local Python setup and install its required attention backend. Its recommended 4× setting is aimed at video super-resolution workflows.
Verdict
FlashVSR provides a self-hosted approach with listed model versions, formats, and batch processing. Hardware compatibility and the required backend are key setup considerations.
Plans and prices
From FlashVSR's own pricing page, checked 9 Oct 2026.
| Plan | Price | What it gives you |
|---|---|---|
| FlashVSR | Free | Open-source code and model weights · Requires a local Python environment and Block-Sparse Attention backend · GPU compatibility varies |
The facts
Each one read from FlashVSR's own pages, with the date it was checked.
What it does
- Community integrations
- The README lists community ComfyUI implementations and WanVideoWrapper integration, and says WanVideoWrapper has known issues.checked 9 Oct 2026
- Core methods
- Its framework combines a three-stage distillation pipeline, locality-constrained sparse attention, and a tiny conditional decoder.checked 9 Oct 2026
- Hardware requirements
- The required Block-Sparse Attention backend is reported to work correctly on NVIDIA A100 and A800, while H200 runs with limited acceleration and compatibility on other listed GPUs is unknown.checked 9 Oct 2026
- Install requirements
- The setup instructions use Python 3.11.13 and require installing the Block-Sparse Attention backend.checked 9 Oct 2026
- License
- The GitHub repository identifies its license as Apache-2.0.checked 9 Oct 2026
- Model versions
- The repository provides v1 and v1.1 model weights through Hugging Face, and recommends v1.1.checked 9 Oct 2026
- Performance
- The project reports about 17 FPS for 768 × 1408 video on a single A100 GPU.checked 9 Oct 2026
- Purpose
- FlashVSR is a one-step diffusion framework for streaming video super-resolution.checked 9 Oct 2026
- Quality limitation
- The project says implementations without its sparse attention module may have noticeable quality degradation, especially at higher resolutions.checked 9 Oct 2026
- Recommended setting
- The project is primarily designed and optimized for 4× video super-resolution, which it recommends for better results and stability.checked 9 Oct 2026
- Support
- The maintainers welcome feedback and issues, and list Junhao Zhuang as a contact.checked 9 Oct 2026
- Third-party inference
- The README lists third-party cloud and API deployments and cautions that their model versions or pipeline completeness may differ from the official implementation.checked 9 Oct 2026
- Training dataset
- VSR-120K is described as containing 120,000 videos and 180,000 images for joint image and video training.checked 9 Oct 2026
Alternatives to FlashVSR
- 1
TensorPix
Free planAndroidiOSWebAPI8.3From $5.50 a month - 2
Cutout.Pro
Free planAndroidiOSWebAPI8.2From $5 a month - 3
Media.io AI Video Upscaler
Free planAndroidiOSWebAPI8.1From $167.88 a year - 4
HitPaw VikPea
AndroidiOSWebAPI7.8From $99.99 a year - 5
Pixop
Free planWebAPILinuxSelf-hosted7.6From $0.02 a month - 6
UniFab Video Enhancer Pro
WebmacOSWindows7.2From $89.99 a year