int 16

app.py

@@ -89,7 +89,7 @@ def response(audio: tuple[int, np.ndarray], conversation: list[dict], img: str |
                             audio_data = frame.split(b'\r\n\r\n', 1)[1]
                             # audio_data = base64.b64decode(audio_data)
                             output_audio_bytes += audio_data
-                            audio_array = np.frombuffer(audio_data, dtype=np.int8).reshape(1, -1)
+                            audio_array = np.frombuffer(audio_data, dtype=np.int16).reshape(1, -1)
                             yield (OUT_RATE, audio_array, "mono")
                         elif b'Content-Type: text/plain' in frame:
                             text_data = frame.split(b'\r\n\r\n', 1)[1].decode()