Update README.md
Browse files
README.md
CHANGED
|
@@ -7,9 +7,7 @@ library_name: transformers
|
|
| 7 |
|
| 8 |
VoiceRestore is a cutting-edge speech restoration model designed to significantly enhance the quality of degraded voice recordings. Leveraging flow-matching transformers, this model excels at addressing a wide range of audio imperfections commonly found in speech, including background noise, reverberation, distortion, and signal loss.
|
| 9 |
|
| 10 |
-
|
| 11 |
-
|
| 12 |
-
Credits: This repository is based on the [E2-TTS implementation by Lucidrains](https://github.com/lucidrains/e2-tts-pytorch)
|
| 13 |
|
| 14 |
## Usage
|
| 15 |
``` bash
|
|
@@ -33,28 +31,23 @@ model("test_input.wav", "test_output.wav")
|
|
| 33 |
## Example
|
| 34 |
### Degraded Input:
|
| 35 |
|
| 36 |
-
|
| 37 |
-
|
| 38 |
-
Degraded audio (reverberation, distortion, noise, random cut):
|
| 39 |
|
| 40 |
-
|
| 41 |
-
|
| 42 |
-
|
|
|
|
| 43 |
|
| 44 |
---
|
| 45 |
### Restored (steps=32, cfg=1.0):
|
| 46 |
|
| 47 |
-
|
|
|
|
|
|
|
|
|
|
| 48 |
|
| 49 |
Restored audio - 16 steps, strength 0.5:
|
| 50 |
|
| 51 |
-
https://github.com/user-attachments/assets/fdbbb988-9bd2-4750-bddd-32bd5153d254
|
| 52 |
-
|
| 53 |
-
---
|
| 54 |
-
### Ground Truth:
|
| 55 |
-
|
| 56 |
-
|
| 57 |
-
|
| 58 |
---
|
| 59 |
## Key Features
|
| 60 |
|
|
@@ -65,7 +58,6 @@ https://github.com/user-attachments/assets/fdbbb988-9bd2-4750-bddd-32bd5153d254
|
|
| 65 |
---
|
| 66 |
|
| 67 |
|
| 68 |
-
|
| 69 |
## Model Details
|
| 70 |
|
| 71 |
- **Architecture**: Flow-matching transformer
|
|
@@ -99,4 +91,5 @@ This project is licensed under the MIT License - see the [LICENSE](LICENSE) file
|
|
| 99 |
## Acknowledgments
|
| 100 |
|
| 101 |
- Based on the [E2-TTS implementation by Lucidrains](https://github.com/lucidrains/e2-tts-pytorch)
|
| 102 |
-
- Special thanks to the open-source community for their invaluable contributions.
|
|
|
|
|
|
| 7 |
|
| 8 |
VoiceRestore is a cutting-edge speech restoration model designed to significantly enhance the quality of degraded voice recordings. Leveraging flow-matching transformers, this model excels at addressing a wide range of audio imperfections commonly found in speech, including background noise, reverberation, distortion, and signal loss.
|
| 9 |
|
| 10 |
+
It is based on this [repo](https://github.com/skirdey/voicerestore) & demo of audio restorations: [VoiceRestore](https://sparkling-rabanadas-3082be.netlify.app/)
|
|
|
|
|
|
|
| 11 |
|
| 12 |
## Usage
|
| 13 |
``` bash
|
|
|
|
| 31 |
## Example
|
| 32 |
### Degraded Input:
|
| 33 |
|
| 34 |
+
### Degraded Input Audio
|
|
|
|
|
|
|
| 35 |
|
| 36 |
+
<audio controls>
|
| 37 |
+
<source src="https://huggingface.co/jadechoghari/VoiceRestore/resolve/main/test_input.wav" type="audio/mpeg">
|
| 38 |
+
Your browser does not support the audio element.
|
| 39 |
+
</audio>
|
| 40 |
|
| 41 |
---
|
| 42 |
### Restored (steps=32, cfg=1.0):
|
| 43 |
|
| 44 |
+
<audio controls>
|
| 45 |
+
<source src="https://huggingface.co/jadechoghari/VoiceRestore/resolve/main/test_output.wav" type="audio/mpeg">
|
| 46 |
+
Your browser does not support the audio element.
|
| 47 |
+
</audio>
|
| 48 |
|
| 49 |
Restored audio - 16 steps, strength 0.5:
|
| 50 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 51 |
---
|
| 52 |
## Key Features
|
| 53 |
|
|
|
|
| 58 |
---
|
| 59 |
|
| 60 |
|
|
|
|
| 61 |
## Model Details
|
| 62 |
|
| 63 |
- **Architecture**: Flow-matching transformer
|
|
|
|
| 91 |
## Acknowledgments
|
| 92 |
|
| 93 |
- Based on the [E2-TTS implementation by Lucidrains](https://github.com/lucidrains/e2-tts-pytorch)
|
| 94 |
+
- Special thanks to the open-source community for their invaluable contributions.
|
| 95 |
+
- Credits: This repository is based on the [E2-TTS implementation by Lucidrains](https://github.com/lucidrains/e2-tts-pytorch)
|