Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -1,3 +1,20 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
# Fish Audio S2 Pro — FP8 (AEmotionStudio Mirror)
|
| 2 |
|
| 3 |
FP8 weight-only quantization of [Fish Audio S2 Pro](https://huggingface.co/fishaudio/s2-pro).
|
|
@@ -24,9 +41,15 @@ in bfloat16. No external quantization library is needed — dequantization is pu
|
|
| 24 |
W_bf16 = W_fp8.to(torch.bfloat16) * scale
|
| 25 |
```
|
| 26 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 27 |
## License
|
| 28 |
|
| 29 |
-
|
| 30 |
-
|
|
|
|
|
|
|
| 31 |
|
| 32 |
Built with Fish Audio.
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: other
|
| 3 |
+
license_name: fish-audio-research
|
| 4 |
+
license_link: LICENSE
|
| 5 |
+
tags:
|
| 6 |
+
- text-to-speech
|
| 7 |
+
- tts
|
| 8 |
+
- voice-cloning
|
| 9 |
+
- speech-synthesis
|
| 10 |
+
- fp8
|
| 11 |
+
- quantized
|
| 12 |
+
language:
|
| 13 |
+
- en
|
| 14 |
+
- zh
|
| 15 |
+
base_model: AEmotionStudio/fish-speech-s2-pro
|
| 16 |
+
---
|
| 17 |
+
|
| 18 |
# Fish Audio S2 Pro — FP8 (AEmotionStudio Mirror)
|
| 19 |
|
| 20 |
FP8 weight-only quantization of [Fish Audio S2 Pro](https://huggingface.co/fishaudio/s2-pro).
|
|
|
|
| 41 |
W_bf16 = W_fp8.to(torch.bfloat16) * scale
|
| 42 |
```
|
| 43 |
|
| 44 |
+
## Usage with ComfyUI-FFMPEGA
|
| 45 |
+
|
| 46 |
+
This model is automatically downloaded and used by the [ComfyUI-FFMPEGA](https://github.com/AEmotionStudio/ComfyUI-FFMPEGA) extension for TTS and voice cloning features when FP8 precision is selected.
|
| 47 |
+
|
| 48 |
## License
|
| 49 |
|
| 50 |
+
**Fish Audio Research License** — see [LICENSE](LICENSE) file.
|
| 51 |
+
|
| 52 |
+
- ✅ Free for research and non-commercial use
|
| 53 |
+
- ❌ Commercial use requires a separate license from [Fish Audio](https://fish.audio) (contact: business@fish.audio)
|
| 54 |
|
| 55 |
Built with Fish Audio.
|