From 9f7ba181207f6b3e1db98d9c3cef7dbb92cd48b9 Mon Sep 17 00:00:00 2001
From: HughhhhCoder <hughz926@gmail.com>
Date: Wed, 2 Sep 2026 11:07:24 +0800
Subject: [PATCH] fix(voice): decode PCM16 output as little-endian

---
 src/agents/voice/result.py   |  4 +++-
 tests/voice/test_pipeline.py | 21 +++++++++++++++++++++
 2 files changed, 24 insertions(+), 1 deletion(-)

diff --git a/src/agents/voice/result.py b/src/agents/voice/result.py
index a01f7d762c..d5d0ecccb8 100644
--- a/src/agents/voice/result.py
+++ b/src/agents/voice/result.py
@@ -104,7 +104,9 @@ def _transform_audio_buffer(
             # np.int16 needs 2-byte alignment; pad odd-length chunks safely.
             combined_buffer += b"\x00"
 
-        np_array = np.frombuffer(combined_buffer, dtype=np.int16)
+        # TTS providers return PCM16 bytes in little-endian order. Decode that wire format
+        # explicitly, then normalize the result to the native int16 dtype returned to callers.
+        np_array = np.frombuffer(combined_buffer, dtype="<i2").astype(np.int16, copy=False)
 
         if output_dtype == np.int16:
             return np_array
diff --git a/tests/voice/test_pipeline.py b/tests/voice/test_pipeline.py
index 02b825376d..ada4a253a9 100644
--- a/tests/voice/test_pipeline.py
+++ b/tests/voice/test_pipeline.py
@@ -73,6 +73,27 @@ def test_streamed_audio_result_odd_length_buffer_int16() -> None:
     assert transformed.tolist() == [1]
 
 
+def test_streamed_audio_result_decodes_pcm16_as_little_endian(monkeypatch) -> None:
+    result = StreamedAudioResult(
+        ZeroPcmTTSModel(),
+        TTSModelSettings(dtype=np.int16),
+        VoicePipelineConfig(),
+    )
+    native_frombuffer = np.frombuffer
+
+    def simulate_big_endian_host(buffer, *, dtype):
+        if dtype is np.int16:
+            dtype = np.dtype(">i2")
+        return native_frombuffer(buffer, dtype=dtype)
+
+    monkeypatch.setattr(np, "frombuffer", simulate_big_endian_host)
+
+    transformed = result._transform_audio_buffer([b"\x01\x02\x03\x04"], np.int16)
+
+    assert transformed.dtype == np.int16
+    assert transformed.tolist() == [0x0201, 0x0403]
+
+
 @pytest.mark.asyncio
 async def test_streamed_audio_result_raises_falsy_task_exception() -> None:
     class FalsyError(RuntimeError):