Update app.py

app.py

@@ -251,20 +251,27 @@ def tts_full(text: str, voice: str = "af_heart", speed: float = 1.0, format: str
 
     # Write the concatenated audio to an in-memory WAV or Opus file.
     sample_rate = 24000
+
+    # Normalize audio data to the range [-1.0, 1.0]
+    audio_normalized = audio / np.max(np.abs(audio))
+
+    # Scale to 16-bit integer range
+    audio_scaled = np.int16(audio_normalized * 32767)
+    
     if format.lower() == "wav":
         
         # Create an in-memory buffer
         wav_io = io.BytesIO()
         
         # Write the audio data to the buffer in WAV format
-        write_wav(wav_io, sample_rate, audio)
+        write_wav(wav_io, sample_rate, audio_scaled)
         
         # Seek to the beginning of the buffer
         wav_io.seek(0)
         
         return Response(content=wav_io.read(), media_type="audio/wav")
     elif format.lower() == "opus":
-        opus_data = audio_tensor_to_opus_bytes(torch.from_numpy(full_audio), sample_rate=sample_rate)
+        opus_data = audio_tensor_to_opus_bytes(torch.from_numpy(audio_scaled), sample_rate=sample_rate)
         return Response(content=opus_data, media_type="audio/opus")
     else:
         raise HTTPException(status_code=400, detail=f"Unsupported audio format: {format}")