app.py

import torch
import soundfile as sf
import gradio as gr
from clearvoice import ClearVoice

myClearVoice = ClearVoice(task='speech_enhancement', model_names=['FRCRN_SE_16K'])

def fn_clearvoice(input_wav):
    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'
  
se_demo = gr.Interface(
    fn=fn_clearvoice,
    inputs = [
        gr.Audio(label="Input Audio", type="filepath")
    ],
    outputs = [
        gr.Audio(label="Output Audio", type="filepath")
    ],
    title = "ClearVoice",
    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/speechbrain/speechbrain' target='_blank'>Github Repo</a></p>"),
    examples = [
        ['mandarin_speech.wav']
    ],
    cache_examples = True,
)
se_demo.launch()