Refactor audio generation in NotebookMg to write directly to output file, eliminating the need for intermediate AudioSegment. This change improves performance and simplifies the audio processing workflow.

main.py

@@ -227,50 +227,45 @@ class NotebookMg:
 
     def generate_audio(self, speaker_lines: List[Tuple[str, str]], output_path: str):
         """Step 4: Generate and stitch audio using ElevenLabs"""
-        combined_audio = AudioSegment.empty()
-
-        for i, (speaker, line) in enumerate(speaker_lines):
-            # Select appropriate voice ID
-            voice_id = (
-                self.Akshara_voice_id if speaker == "Akshara" else self.Tharun_voice_id
-            )
-
-            # Get previous and next text for context
-            previous_text = speaker_lines[i - 1][1] if i > 0 else None
-            next_text = speaker_lines[i + 1][1] if i < len(speaker_lines) - 1 else None
-
-            # Update voice settings with required parameters
-            self.eleven_client.voices.edit_settings(
-                voice_id=voice_id,
-                request=VoiceSettings(
-                    stability=0.5,  # Add stability parameter (0-1)
-                    similarity_boost=0.75,  # Add similarity_boost parameter (0-1)
-                    use_speaker_boost=True,
-                ),
-            )
-
-            # Generate audio for this line
-            audio_data = self.eleven_client.text_to_speech.convert(
-                voice_id=voice_id,
-                output_format="mp3_44100_128",
-                text=line,
-                model_id="eleven_multilingual_v2",
-                previous_text=previous_text,
-                next_text=next_text,
-            )
-
-            # Convert audio data to AudioSegment
-            audio_bytes = b"".join(audio_data)
-            audio_segment = AudioSegment.from_file(
-                io.BytesIO(audio_bytes), format="mp3"
-            )
-
-            # Add small pause between segments
-            pause = AudioSegment.silent(duration=300)  # 300ms pause
-            combined_audio += audio_segment + pause
-
-        # Export final audio
-        combined_audio.export(output_path, format="mp3")
+        # Open the output file in binary write mode
+        with open(output_path, "wb") as outfile:
+            for i, (speaker, line) in enumerate(speaker_lines):
+                # Select appropriate voice ID
+                voice_id = (
+                    self.Akshara_voice_id
+                    if speaker == "Akshara"
+                    else self.Tharun_voice_id
+                )
+
+                # Get previous and next text for context
+                previous_text = speaker_lines[i - 1][1] if i > 0 else None
+                next_text = (
+                    speaker_lines[i + 1][1] if i < len(speaker_lines) - 1 else None
+                )
+
+                # Update voice settings with required parameters
+                self.eleven_client.voices.edit_settings(
+                    voice_id=voice_id,
+                    request=VoiceSettings(
+                        stability=0.5,
+                        similarity_boost=0.75,
+                        use_speaker_boost=True,
+                    ),
+                )
+
+                # Generate audio for this line
+                audio_data = self.eleven_client.text_to_speech.convert(
+                    voice_id=voice_id,
+                    output_format="mp3_44100_128",
+                    text=line,
+                    model_id="eleven_multilingual_v2",
+                    previous_text=previous_text,
+                    next_text=next_text,
+                )
+
+                # Write the audio data directly to the file
+                audio_bytes = b"".join(audio_data)
+                outfile.write(audio_bytes)
 
 
 # def main():