denose_diar

app.py

@@ -4,20 +4,22 @@ from scipy.io.wavfile import write
 from simple_diarizer.diarizer import Diarizer
 from simple_diarizer.utils import (check_wav_16khz_mono, convert_wavfile)
 
+from utils.denoise_pipeline import denoise
 
+def main_pipeline(audio):
+    denoised_audio=denoise(audio)
+    result=denoise(denoised_audio)
 
-def inference(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")
-    return "./out/htdemucs/test/vocals.wav", "./out/htdemucs/test/no_vocals.wav"
-
+    return result
 
 title = "audio_denoise and speakser diarization"
 
 gr.Interface(
-    inference,
+    main_pipeline,
     gr.Audio(type="numpy", label="Input"),
-    [gr.Audio(type="filepath", label="Vocal"), gr.Audio(type="filepath", label="No Vocals / Instrumental")],
+    [
+        gr.Audio(type="filepath", label="Vocal"),
+     # gr.Audio(type="filepath", label="No Vocals / Instrumental")
+    ],
     title=title,
 ).launch(enable_queue=True)

utils/denoise_pipeline.py

@@ -0,0 +1,12 @@
+import os
+import torch
+from scipy.io.wavfile import write
+
+
+def denoise(audio):
+    os.makedirs("out", exist_ok=True)
+    write('test.wav', audio[0], audio[1])
+    device = "cuda:0" if torch.cuda.is_available() else "cpu"
+
+    os.system(f"python3 -m demucs.separate -n htdemucs --two-stems=vocals -d {device} test.wav -o out")
+    return "./out/htdemucs/test/vocals.wav"  # , "./out/htdemucs/test/no_vocals.wav"

utils/diarization_pipeline.py

@@ -4,8 +4,6 @@ from simple_diarizer.utils import (check_wav_16khz_mono, convert_wavfile)
 import soundfile as sf
 
 
-
-
 class DiarizationPipeline:
     def __init__(self, mode='torch'):
         super(DiarizationPipeline, self).__init__()
@@ -22,13 +20,19 @@ class DiarizationPipeline:
 
         # signal, fs = sf.read(wav_file)
         #
-        segments = self.diar.diarize(wav_file,
-                                     num_speakers=None,
-                                     threshold=9e-1,)
+        # segments = self.diar.diarize(wav_file,
+        #                              num_speakers=None,
+        #                              threshold=9e-1, )
+
+        return wav_file
+
+
+pipeline = DiarizationPipeline('torch')
+
+def run_diarization(audio):
 
-        return segments
+    return pipeline(audio)
 
 
 if __name__ == '__main__':
-    pipeline = DiarizationPipeline('torch')
     pipeline('path_audio')