updates

app.py

@@ -26,6 +26,7 @@ def transcribe(aud_inp):
     aud = whisper.pad_or_trim(aud)
     mel = whisper.log_mel_spectrogram(aud).to(device)
     _, probs = WHISPER_MODEL.detect_language(mel)
+    print(f"spectrogram.shape = {mel}")
 
     if device == "cpu":
         options = whisper.DecodingOptions(fp16 = False, language=WHISPER_LANG)
@@ -241,8 +242,8 @@ with tap_input:
         wav_audio_data = st_audiorec()
         if wav_audio_data is not None:
             st.audio(wav_audio_data, format='audio/wav')
-            bytes_to_wav(wav_audio_data, sample_rate=16000)
-            user_input = transcribe(wav_audio_data)
+            bytes_to_wav(wav_audio_data, output_wav_file='output.wav', sample_rate=16000)
+            user_input = transcribe('output.wav')
 
     if submitted or wav_audio_data is not None:
         st.session_state['user_input_content'] = user_input