HawkEye098432's picture
Create app.py
c4bae49
raw
history blame
3.2 kB
import os
import gradio as gr
from scipy.io.wavfile import write
def separate_audio(audio):
os.makedirs("out", exist_ok=True)
write('test.wav', audio[0], audio[1])
os.system("python3 -m demucs.separate -n htdemucs --two-stems=vocals -d cpu test.wav -o out")
vocals_file = "./out/htdemucs/test/vocals.wav"
instrumental_file = "./out/htdemucs/test/no_vocals.wav"
return vocals_file, instrumental_file
def batch_separate_audio(audio_list):
os.makedirs("out", exist_ok=True)
vocals_files = []
instrumental_files = []
for idx, audio in enumerate(audio_list):
write(f'test{idx}.wav', audio[0], audio[1])
os.system(f"python3 -m demucs.separate -n htdemucs --two-stems=vocals -d cpu test{idx}.wav -o out")
vocals_file = f"./out/htdemucs/test{idx}/vocals.wav"
instrumental_file = f"./out/htdemucs/test{idx}/no_vocals.wav"
vocals_files.append(vocals_file)
instrumental_files.append(instrumental_file)
return vocals_files, instrumental_files
def download_file(filepath):
with open(filepath, "rb") as f:
file_bytes = f.read()
return file_bytes
title = "Demucs Music Source Separation (v4)"
description = "This is the latest 'bleeding edge version' which enables the new v4 Hybrid Transformer model. <br> for this space, 2 stem separation (Karaoke Mode) is enabled and CPU mode which has been optimized for best quality & processing time. <p>| Gradio demo for Demucs(v4): Music Source Separation in the Waveform Domain. To use it, simply upload your audio, or click one of the examples to load them. Read more at the links below.</p>"
article = "<p style='text-align: center'><a href='https://arxiv.org/abs/1911.13254' target='_blank'>Music Source Separation in the Waveform Domain</a> | <a href='https://github.com/facebookresearch/demucs' target='_blank'>Github Repo</a> | <a href='https://www.thafx.com' target='_blank'>//THAFX</a></p>"
audio_input = gr.inputs.Audio(label="Input")
vocals_output = gr.outputs.Audio(label="Vocals", type="filepath", download=True)
instrumental_output = gr.outputs.Audio(label="No Vocals / Instrumental", type="filepath", download=True)
examples = [['test.mp3']]
# Create the Gradio interface
gr.Interface(
fn=separate_audio,
inputs=audio_input,
outputs=[vocals_output, instrumental_output],
title=title,
description=description,
article=article,
examples=examples
).launch(enable_queue=True, share=True)
batch_audio_input = gr.inputs.Audio(label="Input", type="numpy", multiple=True)
batch_vocals_output = gr.outputs.Audio(label="Vocals", type="filepath", download=True, multiple=True)
batch_instrumental_output = gr.outputs.Audio(label="No Vocals / Instrumental", type="filepath", download=True, multiple=True)
batch_examples = [[audio] for audio in examples[0]]
# Create the Gradio interface for batch conversion
gr.Interface(
fn=batch_separate_audio,
inputs=batch_audio_input,
outputs=[batch_vocals_output, batch_instrumental_output],
title="Demucs Batch Music Source Separation (v4)",
description=description,
article=article,
examples=batch_examples
).launch(enable_queue=True, share=True)