Update app.py

app.py

@@ -1,22 +1,24 @@
-#
 import streamlit as st
 from tempfile import NamedTemporaryFile
 from audiorecorder import audiorecorder
 from whispercpp import Whisper
+from pydub import AudioSegment
+import io
 
-# Download whisper.cpp
+# Initialize whisper.cpp
 w = Whisper('tiny')
 
-def inference(audio_data):
-    # Save the raw audio data to a temporary file
+def inference(audio_segment):
+    # Convert AudioSegment to WAV format in memory
     with NamedTemporaryFile(suffix=".wav", delete=False) as temp:
-        temp.write(audio_data)  # write the raw audio bytes
+        # Export AudioSegment to raw bytes in WAV format
+        audio_segment.export(temp.name, format="wav")
         temp.close()  # Ensure the file is written and closed before passing it to Whisper
         result = w.transcribe(temp.name)
         text = w.extract_text(result)
     return text[0]
 
-# Streamlit
+# Streamlit UI setup
 with st.sidebar:
     audio = audiorecorder("Click to send voice message", "Recording... Click when you're done", key="recorder")
     st.title("Echo Bot with Whisper")
@@ -34,7 +36,9 @@ for message in st.session_state.messages:
 if (prompt := st.chat_input("Your message")) or len(audio):
     # If it's coming from the audio recorder transcribe the message with whisper.cpp
     if len(audio) > 0:
-        prompt = inference(audio)
+        # Convert audio (from audiorecorder) to AudioSegment
+        audio_segment = AudioSegment.from_file(io.BytesIO(audio), format="wav")
+        prompt = inference(audio_segment)
 
     # Display user message in chat message container
     st.chat_message("user").markdown(prompt)