import gradio as gr from audio_style_transfer.models import timedomain import librosa def audioStyleTransfer(content,style): print (style,content) output = "/tmp/outputfile.wav" sr = librosa.get_samplerate(content) timedomain.run( style_fname=style, content_fname=content ,output_fname=output, n_fft=4096, n_filters=4096, hop_length=256, alpha=0.05, k_w=4, # 1 to 3. Lower is better quality but is slower. sr=sr) print ("output is " ,output) return output iface = gr.Interface(fn=audioStyleTransfer, title="Time Domain Audio Style transfer", description="Forked from https://github.com/pkmital/time-domain-neural-audio-style-transfer Built to style transfer to audio using style audio.\ it seems to work best for shorter clips", inputs=[gr.Audio(source="upload",type="filepath",label="Content"),gr.Audio(source="upload",type="filepath",label="Style")], outputs=gr.Audio(label="Output")) iface.launch()