File size: 1,633 Bytes
06fc5c8 00885ea 0075f67 00885ea 0075f67 00885ea 0075f67 00885ea e34c54b 00885ea 0075f67 00885ea 06fc5c8 e34c54b 06fc5c8 0075f67 00885ea 06fc5c8 00885ea 8581ee6 06fc5c8 0075f67 |
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 |
import os
import time
import json
import gradio as gr
import torch
import torchaudio
import numpy as np
from denoiser.demucs import Demucs
from pydub import AudioSegment
modelpath = './denoiser/master64.th'
def transcribe(file_upload, microphone):
file = microphone if microphone is not None else file_upload
model = Demucs(hidden=64)
state_dict = torch.load(modelpath, map_location='cpu')
model.load_state_dict(state_dict)
demucs = model
x, sr = torchaudio.load(file)
out = demucs(x[None])[0]
out = out / max(out.abs().max().item(), 1)
torchaudio.save('enhanced.wav', out, sr)
enhanced = AudioSegment.from_wav('enhanced.wav') # 只有去完噪的需要降 bitrate 再做語音識別
enhanced.export('enhanced.wav', format="wav", bitrate="256k")
return "enhanced.wav"
demo = gr.Interface(
fn=transcribe,
inputs=[
gr.Audio(type="filepath", label="語音質檢麥克風實時錄音"),
gr.Audio(type="filepath", label="語音質檢原始音檔"),
],
outputs=gr.Audio(type="filepath", label="Output"),
title="<p style='text-align: center'><a href='https://www.twman.org/AI' target='_blank'>語音質檢噪音去除 (語音增強)：Meta Denoiser</a>",
description="為了提升語音識別的效果，可以在識別前先進行噪音去除",
allow_flagging="never",
examples=[
["exampleAudio/15s_2020-03-27_sep1.wav"],
["exampleAudio/13s_2020-03-27_sep2.wav"],
["exampleAudio/30s_2020-04-23_sep1.wav"],
["exampleAudio/15s_2020-04-23_sep2.wav"],
],
)
demo.launch(enable_queue=True, debug=True) |