from speechbrain.pretrained import SepformerSeparation as separator import torchaudio import gradio as gr model = separator.from_hparams(source="speechbrain/sepformer-wsj02mix", savedir='pretrained_models\sepformer-wsj02mix') def speechbrain(aud): est_sources = model.separate_file(path=aud) torchaudio.save("source1hat.wav", est_sources[:, :, 0].detach().cpu(), 8000) torchaudio.save("source2hat.wav", est_sources[:, :, 1].detach().cpu(), 8000) return "source1hat.wav", "source2hat.wav" inputs = gr.components.Audio(label="Input Audio", type="filepath") outputs = [ gr.components.Audio(label="Output Audio One", type="filepath"), gr.components.Audio(label="Output Audio Two", type="filepath") ] title = "Speech Seperation" description = "Gradio demo for Speech Seperation by SpeechBrain. To use it, simply upload your audio, or click one of the examples to load them. Read more at the links below." article = "

Attention is All You Need in Speech Separation | Github Repo

" examples = [ ['samples_audio_samples_test_mixture.wav'] ] gr.Interface(speechbrain, inputs, outputs, title=title, description=description, article=article, examples=examples).launch()