Spaces:
Running
on
Zero
Running
on
Zero
File size: 3,667 Bytes
02c7bdf bdaf47a 02c7bdf b78b7d0 a9e592e 132a2a9 f821359 02c7bdf 3192961 b78b7d0 4554491 b78b7d0 4554491 b78b7d0 3192961 3956066 b78b7d0 e805751 b78b7d0 e805751 b78b7d0 6963e61 b78b7d0 6320c59 b78b7d0 8bb6908 3956066 e805751 3956066 e805751 3956066 b78b7d0 3956066 6963e61 3956066 6320c59 d04ae35 3956066 b78b7d0 99710ec 6963e61 3acf4c8 0f8dddd |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 |
import torch
import soundfile as sf
import gradio as gr
from clearvoice import ClearVoice
def fn_clearvoice_se(input_wav):
myClearVoice = ClearVoice(task='speech_enhancement', model_names=['FRCRN_SE_16K'])
output_wav_dict = myClearVoice(input_path=input_wav, online_write=False)
if isinstance(output_wav_dict, dict):
key = next(iter(output_wav_dict))
output_wav = output_wav_dict[key]
else:
output_wav = output_wav_dict
sf.write('enhanced.wav', output_wav, 16000)
return 'enhanced.wav'
def fn_clearvoice_ss(input_wav):
myClearVoice = ClearVoice(task='speech_separation', model_names=['MossFormer2_SS_16K'])
output_wav_dict = myClearVoice(input_path=input_wav, online_write=False)
if isinstance(output_wav_dict, dict):
key = next(iter(output_wav_dict))
output_wav_list = output_wav_dict[key]
output_wav_s1 = output_wav_list[0]
output_wav_s2 = output_wav_list[1]
else:
output_wav_list = output_wav_dict
output_wav_s1 = output_wav_list[0]
output_wav_s2 = output_wav_list[1]
sf.write('separated_s1.wav', output_wav_s1, 16000)
sf.write('separated_s2.wav', output_wav_s2, 16000)
return "separated_s1.wav", "separated_s2.wav"
demo = gr.Blocks()
se_demo = gr.Interface(
fn=fn_clearvoice_se,
inputs = [
gr.Audio(label="Input Audio", type="filepath"),
],
outputs = [
gr.Audio(label="Output Audio", type="filepath"),
],
title = "ClearVoice: Speech Enhancement",
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."),
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/alibabasglab/FRCRN' target='_blank'>Github Repo</a></p>"
),
examples = [
['examples/mandarin_speech_16kHz.wav'],
['examples/english_speech_48kHz.wav'],
],
cache_examples = True,
)
ss_demo = gr.Interface(
fn=fn_clearvoice_ss,
inputs = [
gr.Audio(label="Input Audio", type="filepath"),
],
outputs = [
gr.Audio(label="Output Audio", type="filepath"),
gr.Audio(label="Output Audio", type="filepath"),
],
title = "ClearVoice: Speech Separation",
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."),
article = ("<p style='text-align: center'><a href='https://arxiv.org/abs/2302.11824' target='_blank'>MossFormer: Pushing the Performance Limit of Monaural Speech Separation using Gated Single-Head Transformer with Convolution-Augmented Joint Self-Attentions</a> | <a href='https://github.com/alibabasglab/MossFormer' target='_blank'>Github Repo</a></p>"
"<p style='text-align: center'><a href='https://arxiv.org/abs/2312.11825' target='_blank'>MossFormer2: Combining Transformer and RNN-Free Recurrent Network for Enhanced Time-Domain Monaural Speech Separation</a> | <a href='https://github.com/alibabasglab/MossFormer2' target='_blank'>Github Repo</a></p>"),
examples = [
['examples/female_female_speech.wav'],
['examples/female_male_speech.wav'],
],
cache_examples = True,
)
with demo:
#gr.TabbedInterface([se_demo], ["Speech Enhancement"])
gr.TabbedInterface([se_demo, ss_demo], ["Speech Enhancement", "Speech Separation"])
demo.launch() |