Update app.py

app.py

@@ -5,6 +5,10 @@ import time
 from transformers import pipeline
 from pydub import AudioSegment
 import speech_recognition as sr
+from streamlit_webrtc import webrtc_streamer, VideoTransformerBase
+import numpy as np
+import wave
+import io
 
 # Streamlit Configuration
 st.set_page_config(page_title="Interview Copilot", layout="wide")
@@ -63,39 +67,44 @@ if st.button("Start Interview"):
         if 'responses' not in st.session_state:
             st.session_state.responses = []
 
-        # Step 4: Question and Voice Response
+        # Step 4: Question and Live Voice Response
         current_question = all_questions[st.session_state.question_index]
         st.write(f"**Question {st.session_state.question_index + 1}:** {current_question}")
-        st.write("🎤 Please record your answer and upload the audio file.")
-
-        audio_file = st.file_uploader(f"Upload your answer to question {st.session_state.question_index + 1}:", type=["mp3", "wav", "ogg", "flac"])
+        st.write("🎤 Please record your answer using the microphone.")
         
-        if audio_file:
-            # Process the audio file
-            audio = AudioSegment.from_file(audio_file)
-            audio_path = f"audio_{st.session_state.question_index}.wav"
-            audio.export(audio_path, format="wav")
+        # Live voice recording with streamlit-webrtc
+        audio_recorder = webrtc_streamer(
+            key="audio-recorder", 
+            video_transformer_factory=VideoTransformerBase,
+            audio_source=True,  # Enable audio recording
+            sendback_audio=True  # Send back recorded audio
+        )
+
+        # Check if the user has recorded audio
+        if audio_recorder and audio_recorder.audio:
+            audio_data = audio_recorder.audio
+            audio_bytes = audio_data.tobytes()  # Convert audio to byte format for processing
+            audio_file = io.BytesIO(audio_bytes)  # Use BytesIO to simulate file for audio processing
 
-            # Transcribe audio to text
+            # Process the audio
             recognizer = sr.Recognizer()
-            with sr.AudioFile(audio_path) as source:
-                audio_data = recognizer.record(source)
             try:
-                # Transcribing the speech to text
-                transcript = recognizer.recognize_google(audio_data)
-                st.session_state.responses.append(transcript)
-                st.write(f"📝 Transcribed Response: {transcript}")
-                os.remove(audio_path)  # Clean up the audio file
+                with sr.AudioFile(audio_file) as source:
+                    audio_data = recognizer.record(source)
+                    # Transcribe audio to text
+                    transcript = recognizer.recognize_google(audio_data)
+                    st.session_state.responses.append(transcript)
+                    st.write(f"📝 Transcribed Response: {transcript}")
 
-                # Move to the next question after the answer
-                if st.session_state.question_index < len(all_questions) - 1:
-                    st.session_state.question_index += 1
-                else:
-                    st.session_state.question_index = 0  # Reset to first question after the last one
+                    # Proceed to next question or complete interview
+                    if st.session_state.question_index < len(all_questions) - 1:
+                        st.session_state.question_index += 1
+                    else:
+                        st.session_state.question_index = 0  # Reset to first question after the last one
 
-                # Show the next question or finish the interview
-                if len(st.session_state.responses) < len(all_questions):
-                    st.experimental_rerun()
+                    # Show the next question or finish the interview
+                    if len(st.session_state.responses) < len(all_questions):
+                        st.experimental_rerun()
 
             except sr.UnknownValueError:
                 st.error("Sorry, I couldn't understand the audio.")