test

app.py

@@ -5,10 +5,8 @@ import sys
 sys.path.append('.')
 import gradio as gr
 os.system('pip install -U torchtext==0.8.0')
-os.system('pip install pydub')  # 添加对pydub库的依赖
 #os.system('python setup.py install --install-dir .')
 from scipy.io import wavfile
-from pydub import AudioSegment
 
 os.system('chmod a+x ./separate_scripts/*.sh')
 os.system('chmod a+x ./scripts/*.sh')
@@ -16,19 +14,14 @@ os.system('chmod a+x ./scripts/*/*.sh')
 os.system('./separate_scripts/download_checkpoints.sh')
 
 def inference(audio):
-    # convert mp3 to wav if needed
-    if audio.name.endswith(".mp3"):
-        audio_wav = audio.name[:-4] + ".wav"
-        AudioSegment.from_file(audio.name).export(audio_wav, format="wav")
-        audio.name = audio_wav
     # read the file and get the sample rate and data
     rate, data = wavfile.read(audio.name) 
     
     # save the result
-    wavfile.write('foo_left.wav', rate, data)
-    os.system("""python bytesep/inference.py --config_yaml=./scripts/4_train/musdb18/configs/vocals-accompaniment,resunet_subbandtime.yaml --checkpoint_path=./downloaded_checkpoints/resunet143_subbtandtime_vocals_8.8dB_350k_steps.pth --audio_path=foo_left.wav --output_path=sep_vocals.mp3""")
-    #os.system('./separate_scripts/separate_vocals.sh ' + audio.name + ' "sep_vocals.mp3"')
-    os.system("""python bytesep/inference.py --config_yaml=./scripts/4_train/musdb18/configs/accompaniment-vocals,resunet_subbandtime.yaml --checkpoint_path=./downloaded_checkpoints/resunet143_subbtandtime_accompaniment_16.4dB_350k_steps.pth --audio_path=foo_left.wav --output_path=sep_accompaniment.mp3""")
+    filename = 'foo_left' + os.path.splitext(audio.name)[-1] # 构造文件名，使用原文件名的后缀
+    wavfile.write(filename, rate, data)
+    os.system(f"""python bytesep/inference.py --config_yaml=./scripts/4_train/musdb18/configs/vocals-accompaniment,resunet_subbandtime.yaml --checkpoint_path=./downloaded_checkpoints/resunet143_subbtandtime_vocals_8.8dB_350k_steps.pth --audio_path={filename} --output_path=sep_vocals.mp3""")
+    os.system(f"""python bytesep/inference.py --config_yaml=./scripts/4_train/musdb18/configs/accompaniment-vocals,resunet_subbandtime.yaml --checkpoint_path=./downloaded_checkpoints/resunet143_subbtandtime_accompaniment_16.4dB_350k_steps.pth --audio_path={filename} --output_path=sep_accompaniment.mp3""")
     #os.system('./separate_scripts/separate_accompaniment.sh ' + audio.name + ' "sep_accompaniment.mp3"')
     #os.system('python separate_scripts/separate.py --audio_path=' +audio.name+' --source_type="accompaniment"')
     #os.system('python separate_scripts/separate.py --audio_path=' +audio.name+' --source_type="vocals"')