Update app.py

app.py

@@ -9,11 +9,14 @@ import io
 import glob
 import shutil
 import time
+from hf_transcriber import HFTranscriber
+from transcriber import AudioTranscriber
 
 # Configuration dictionary to store app settings
 app_config = {
     'RECORDING_ENABLED': False,
-    'AUDIO_DEVICES': []
+    'AUDIO_DEVICES': [],
+    'DEFAULT_MODEL': 'openai/whisper-small'  # Default model for transcription
 }
 
 def init_recording():
@@ -318,33 +321,59 @@ def main():
     if 'last_recorded_audio' in st.session_state and st.session_state.last_recorded_audio:
         audio_file = st.session_state.last_recorded_audio
         
+        # Add model selection
+        model_options = {
+            "Whisper Small": "openai/whisper-small",
+            "Whisper Base": "openai/whisper-base",
+            "Wav2Vec2 Base": "facebook/wav2vec2-base-960h",
+            "SpeechT5": "microsoft/speecht5_asr"
+        }
+        selected_model = st.selectbox(
+            "Select Transcription Model",
+            options=list(model_options.keys()),
+            index=0
+        )
+        
         if st.button("🎵 Transcribe Audio"):
             try:
                 with st.spinner("Transcribing audio..."):
-                    # Initialize the transcriber
-                    if use_hf:
-                        transcriber = HFTranscriber(model_name=model_name)
-                    else:
-                        transcriber = AudioTranscriber()
-                
+                    # Initialize the transcriber with the selected model
+                    model_name = model_options[selected_model]
+                    transcriber = HFTranscriber(model_name=model_name)
+                    
+                    # Create output directory
+                    os.makedirs("outputs", exist_ok=True)
+                    timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
+                    
                     try:
-                        # Create output directory
-                        os.makedirs("outputs", exist_ok=True)
-                        timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
-                        midi_output = os.path.join("outputs", f"output_{timestamp}.mid")
-                        musicxml_output = os.path.join("outputs", f"output_{timestamp}.musicxml")
+                        # Transcribe the audio
+                        result = transcriber.transcribe_audio(audio_file)
                         
-                        # Transcribe
-                        transcriber.transcribe(audio_file, midi_output, musicxml_output)
+                        # Save transcription to file
+                        txt_output = os.path.join("outputs", f"transcription_{timestamp}.txt")
+                        with open(txt_output, 'w', encoding='utf-8') as f:
+                            f.write(result['transcription'])
                         
                         # Show success message
                         st.success("🎵 Transcription completed successfully!")
                         
-                        # Show download links
+                        # Display the transcription
+                        st.markdown("### Transcription Result")
+                        st.text_area("Transcription", result['transcription'], height=200)
+                        
+                        # Show download link
                         st.markdown("### Download Results")
+                        st.download_button(
+                            label="Download Transcription",
+                            data=result['transcription'],
+                            file_name=f"transcription_{timestamp}.txt",
+                            mime="text/plain"
+                        )
                         
-                        # Display MIDI file
-                        if os.path.exists(midi_output):
+                        # If we have timing information, show it
+                        if 'word_timestamps' in result and result['word_timestamps']:
+                            st.markdown("### Word-level Timestamps")
+                            st.json(result['word_timestamps'])
                             st.markdown(f"**MIDI File:** {get_binary_file_downloader_html(midi_output, 'Download MIDI')}", 
                                       unsafe_allow_html=True)