Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
@@ -6,28 +6,29 @@ from clearvoice import ClearVoice
|
|
6 |
myClearVoice = ClearVoice(task='speech_enhancement', model_names=['FRCRN_SE_16K'])
|
7 |
|
8 |
def fn_clearvoice(aud):
|
9 |
-
# Load and add fake batch dimension
|
10 |
-
"""
|
11 |
-
noisy = enhance_model.load_audio(
|
12 |
-
aud
|
13 |
-
).unsqueeze(0)
|
14 |
-
enhanced = enhance_model.enhance_batch(noisy, lengths=torch.tensor([1.]))
|
15 |
-
"""
|
16 |
output_wav_dict = myClearVoice(input_path='input.wav', online_write=False)
|
17 |
if isinstance(output_wav_dict, dict):
|
18 |
key = next(iter(output_wav_dict))
|
19 |
output_wav = output_wav_dict[key]
|
20 |
else:
|
21 |
output_wav = output_wav_dict
|
22 |
-
sf.write('enhanced.wav', output_wav, 16000)
|
23 |
return 'enhanced.wav'
|
24 |
|
25 |
-
|
26 |
-
|
27 |
-
|
28 |
-
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
-
]
|
33 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
6 |
myClearVoice = ClearVoice(task='speech_enhancement', model_names=['FRCRN_SE_16K'])
|
7 |
|
8 |
def fn_clearvoice(aud):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
9 |
output_wav_dict = myClearVoice(input_path='input.wav', online_write=False)
|
10 |
if isinstance(output_wav_dict, dict):
|
11 |
key = next(iter(output_wav_dict))
|
12 |
output_wav = output_wav_dict[key]
|
13 |
else:
|
14 |
output_wav = output_wav_dict
|
15 |
+
#sf.write('enhanced.wav', output_wav, 16000)
|
16 |
return 'enhanced.wav'
|
17 |
|
18 |
+
se_demo = gr.Interface(
|
19 |
+
fn=fn_clearvoice,
|
20 |
+
inputs = [
|
21 |
+
gr.Audio(sources=["upload"], label="Input Audio", type="filepath")
|
22 |
+
],
|
23 |
+
outputs = [
|
24 |
+
gr.Audio(label="Output Audio", type="filepath")
|
25 |
+
],
|
26 |
+
title = "ClearVoice",
|
27 |
+
description = ("Gradio demo for Speech enhancement with ClearVoice. To use it, simply upload your audio, or click one of the examples to load them. Read more at the links below."),
|
28 |
+
article = ("<p style='text-align: center'><a href='https://arxiv.org/abs/2206.07293' target='_blank'>FRCRN: Boosting Feature Representation Using Frequency Recurrence for Monaural Speech Enhancement</a> | <a href='https://github.com/speechbrain/speechbrain' target='_blank'>Github Repo</a></p>"),
|
29 |
+
examples = [
|
30 |
+
['input.wav']
|
31 |
+
]
|
32 |
+
cache_examples = True,
|
33 |
+
)
|
34 |
+
se_demo.launch()
|