22 lines
1.2 KiB
Markdown
22 lines
1.2 KiB
Markdown
# ComfyUI-FlashVSR Update Log
|
|
|
|
## V1.1.1 (2025/11/17)
|
|
- Refresh FlashVSR 1.1 main model package (renamed to `FlashVSR1_1.safetensors` alongside `LQ_proj_in.safetensors` and `TCDecoder.safetensors`) and update latest HuggingFace models.
|
|
- Fix the core loader to validate the new filenames and raise clear warnings when any required asset is missing, preventing silent fallbacks.
|
|
- Fix runtime RoPE dtype patch so FlashVSR runs on Apple Silicon/MPS without float64 errors. https://github.com/1038lab/ComfyUI-FlashVSR/issues/7
|
|
|
|
## V1.1.0 (2025/11/15)
|
|
|
|
### 🔧 Fixes
|
|
- Applied frame-duplication fix contributed by **chris87423** and implemented by **goofyrodent**.
|
|
- Removes padded frames at the end of the sequence
|
|
- Removes the duplicated first 2 frames produced by FlashVSR
|
|
- Fix applied to both Basic and Advanced nodes
|
|
- Resolves issue [#3](https://github.com/1038lab/ComfyUI-FlashVSR/issues/3)
|
|
|
|
### 🧩 Model Updates
|
|
- Updated FlashVSR 1.1 model:
|
|
**[Wan2_1-T2V-1.1_3B_FlashVSR_fp32.safetensors](https://huggingface.co/1038lab/FlashVSR/blob/main/Wan2_1-T2V-1.1_3B_FlashVSR_fp32.safetensors)**
|
|
|
|
This model improves T2V → VSR sharpness, detail preservation, and temporal stability.
|