Update app.py

app.py

@@ -12,6 +12,7 @@ import torchaudio
 from pydub import AudioSegment
 import io
 import librosa
+from torchaudio.io import CodecConfig
 # import numpy
 
 # def my_read_file(audio_path, max_second):
@@ -105,9 +106,14 @@ def main():
             wav= wav.unsqueeze(0)
                 
         elif file_extension == ".mp3":
-            torchaudio.set_audio_backend("ffmpeg")
+
+            # Switch to FFmpeg backend
+            torchaudio.set_audio_backend("sox")
+
+            # Define CodecConfig for MP3 compression
+            #codec_config = CodecConfig(format="mp3", compression=128)  # 128 kbps for MP3
             #wav3, sample_rate = librosa.load("test.mp3")
-            wav3, sample_rate = torchaudio.load("test.mp3")
+            wav3, sample_rate = torchaudio.load("test.mp3",compression =-4.5)
             st.markdown(wav3)
             #RuntimeError: Could not infer dtype of numpy.float32
             #wav = torch.tensor(wav3).float() / 32768.0