Add README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,49 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# FlashVSR Model Mirror (AEmotionStudio)
|
| 2 |
+
|
| 3 |
+
Mirror of [FlashVSR v1.1](https://github.com/OpenImagingLab/FlashVSR) model weights for [ComfyUI-FFMPEGA](https://github.com/AEmotionStudio/ComfyUI-FFMPEGA).
|
| 4 |
+
|
| 5 |
+
## Model Files
|
| 6 |
+
|
| 7 |
+
| File | Description | Size |
|
| 8 |
+
|---|---|---|
|
| 9 |
+
| `FlashVSR1_1.safetensors` | DiT v1.1 weights (one-step diffusion backbone) | ~5.7 GB |
|
| 10 |
+
| `Wan2.1_VAE.safetensors` | WAV2.1 VAE decoder | ~254 MB |
|
| 11 |
+
| `LQ_proj_in.safetensors` | Low-quality 4× projection encoder | ~576 MB |
|
| 12 |
+
| `TCDecoder.safetensors` | Tiny Conditional Decoder (for Tiny/Tiny-Long modes) | ~181 MB |
|
| 13 |
+
| `Prompt.safetensors` | Cross-attention conditioning embeddings | ~4 MB |
|
| 14 |
+
|
| 15 |
+
## Variants
|
| 16 |
+
|
| 17 |
+
- **FlashVSR Full** — Best quality, full VAE decode (~12-16 GB VRAM)
|
| 18 |
+
- **FlashVSR Tiny** — Fast mode with TCDecoder (~8-12 GB VRAM)
|
| 19 |
+
- **FlashVSR Tiny Long** — Streaming for long videos, lowest VRAM (~8-12 GB)
|
| 20 |
+
|
| 21 |
+
All variants share the same weight files. The variant determines which pipeline and decoder components are used at inference time.
|
| 22 |
+
|
| 23 |
+
## Usage
|
| 24 |
+
|
| 25 |
+
These models are automatically downloaded by [ComfyUI-FFMPEGA](https://github.com/AEmotionStudio/ComfyUI-FFMPEGA) when selecting any `flashvsr_*` model from the `upscale_model` dropdown.
|
| 26 |
+
|
| 27 |
+
Manual installation:
|
| 28 |
+
```
|
| 29 |
+
ComfyUI/models/FlashVSR/
|
| 30 |
+
├── FlashVSR1_1.safetensors
|
| 31 |
+
├── Wan2.1_VAE.safetensors
|
| 32 |
+
├── LQ_proj_in.safetensors
|
| 33 |
+
├── TCDecoder.safetensors
|
| 34 |
+
└── Prompt.safetensors
|
| 35 |
+
```
|
| 36 |
+
|
| 37 |
+
## Attribution
|
| 38 |
+
|
| 39 |
+
- **Original model**: [FlashVSR (CVPR 2026)](https://github.com/OpenImagingLab/FlashVSR) by Junhao Zhuang et al.
|
| 40 |
+
- **Original weights**: [JunhaoZhuang/FlashVSR](https://huggingface.co/JunhaoZhuang/FlashVSR)
|
| 41 |
+
- **ComfyUI adaptation (BSA-free)**: [1038lab/ComfyUI-FlashVSR](https://github.com/1038lab/ComfyUI-FlashVSR)
|
| 42 |
+
- **ComfyUI integration weights**: [1038lab/FlashVSR](https://huggingface.co/1038lab/FlashVSR)
|
| 43 |
+
|
| 44 |
+
## License
|
| 45 |
+
|
| 46 |
+
GPL-3.0 — See [LICENSE](LICENSE) for full text.
|
| 47 |
+
|
| 48 |
+
The FlashVSR model weights are released under the GNU General Public License v3.0.
|
| 49 |
+
All usage must comply with the original license terms from the FlashVSR authors.
|