added description
Browse files
app.py
CHANGED
|
@@ -39,10 +39,17 @@ def main(wav_path,transcript,lang_code):
|
|
| 39 |
torchaudio.save(fp,cleaned_wav.view(1,-1), sample_rate=22050,format='wav')
|
| 40 |
return fp.name
|
| 41 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 42 |
inputs = [gr.Audio(label="noisy audio",type='filepath'),gr.Textbox(label="Transcript", value="Your transcript here", max_lines=1),
|
| 43 |
gr.Radio(label="Language", choices=["eng-us", "jpn"], value="eng-us")]
|
| 44 |
outputs = gr.Audio(label="Output")
|
| 45 |
|
| 46 |
-
demo = gr.Interface(fn=main, inputs=inputs, outputs=outputs)
|
| 47 |
|
| 48 |
demo.launch()
|
|
|
|
| 39 |
torchaudio.save(fp,cleaned_wav.view(1,-1), sample_rate=22050,format='wav')
|
| 40 |
return fp.name
|
| 41 |
|
| 42 |
+
description = """
|
| 43 |
+
# Miipher demo
|
| 44 |
+
This repository provices pretrained weights and demo of Miipher implementation by [Wataru-Nakata](https://github.com/Wataru-Nakata/miipher)
|
| 45 |
+
Miipher was originally proposed by Koizumi et. al. [arxiv](https://arxiv.org/abs/2303.01664)
|
| 46 |
+
Please note that the model differs in many ways from the paper.
|
| 47 |
+
**Non commercial use only** as the weights are provided in CC-BY-NC 2.0.
|
| 48 |
+
"""
|
| 49 |
inputs = [gr.Audio(label="noisy audio",type='filepath'),gr.Textbox(label="Transcript", value="Your transcript here", max_lines=1),
|
| 50 |
gr.Radio(label="Language", choices=["eng-us", "jpn"], value="eng-us")]
|
| 51 |
outputs = gr.Audio(label="Output")
|
| 52 |
|
| 53 |
+
demo = gr.Interface(fn=main, inputs=inputs, outputs=outputs,description=description)
|
| 54 |
|
| 55 |
demo.launch()
|