File size: 1,260 Bytes
740a68f 8a51868 740a68f 8a51868 740a68f |
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 |
import torch
import torchaudio
from sgmse.model import ScoreModel
import gradio as gr
# Load the pre-trained model
model = ScoreModel.load_from_checkpoint("path/to/your/checkpoint.ckpt")
def enhance_speech(audio_file):
# Load and process the audio file
noisy, sr = torchaudio.load(audio_file)
noisy = noisy.unsqueeze(0) # Add fake batch dimension if needed
# Run the speech enhancement model
enhanced = model.predict(noisy)
# Save the enhanced audio
output_file = 'enhanced_output.wav'
torchaudio.save(output_file, enhanced.cpu().squeeze(0), sr)
return output_file
# Gradio interface setup
inputs = gr.Audio(label="Input Audio", type="filepath")
outputs = gr.Audio(label="Output Audio", type="filepath")
title = "Speech Enhancement using SGMSE"
description = "This Gradio demo uses the SGMSE model for speech enhancement. Upload your audio file to enhance it."
article = "<p style='text-align: center'><a href='https://huggingface.co/SP-UHH/speech-enhancement-sgmse' target='_blank'>Model Card</a></p>"
examples = [
['samples/your_example_audio.wav']
]
gr.Interface(fn=enhance_speech, inputs=inputs, outputs=outputs, title=title, description=description, article=article, examples=examples).launch()
|