TTS to always stream when available (#148695)

Co-authored-by: Michael Hansen <mike@rhasspy.org>

homeassistant/components/tts/__init__.py

@@ -382,7 +382,7 @@ async def _async_convert_audio(
             assert process.stderr
             stderr_data = await process.stderr.read()
             _LOGGER.error(stderr_data.decode())
-            raise RuntimeError(
+            raise HomeAssistantError(
                 f"Unexpected error while running ffmpeg with arguments: {command}. "
                 "See log for details."
             )
@@ -976,7 +976,7 @@ class SpeechManager:
         if engine_instance.name is None or engine_instance.name is UNDEFINED:
             raise HomeAssistantError("TTS engine name is not set.")
 
-        if isinstance(engine_instance, Provider) or isinstance(message_or_stream, str):
+        if isinstance(engine_instance, Provider):
             if isinstance(message_or_stream, str):
                 message = message_or_stream
             else:
@@ -996,8 +996,18 @@ class SpeechManager:
             data_gen = make_data_generator(data)
 
         else:
+            if isinstance(message_or_stream, str):
+
+                async def gen_stream() -> AsyncGenerator[str]:
+                    yield message_or_stream
+
+                stream = gen_stream()
+
+            else:
+                stream = message_or_stream
+
             tts_result = await engine_instance.internal_async_stream_tts_audio(
-                TTSAudioRequest(language, options, message_or_stream)
+                TTSAudioRequest(language, options, stream)
             )
             extension = tts_result.extension
             data_gen = tts_result.data_gen

tests/components/assist_pipeline/snapshots/test_pipeline.ambr

@@ -1,5 +1,5 @@
 # serializer version: 1
-# name: test_chat_log_tts_streaming[to_stream_deltas0-0-]
+# name: test_chat_log_tts_streaming[to_stream_deltas0-1-hello, how are you?]
   list([
     dict({
       'data': dict({

tests/components/assist_pipeline/test_pipeline.py

@@ -1550,9 +1550,9 @@ async def test_pipeline_language_used_instead_of_conversation_language(
                     "?",
                 ],
             ),
-            # We are not streaming, so 0 chunks via streaming method
+            # We always stream when possible, so 1 chunk via streaming method
-            0,
+            1,
-            "",
+            "hello, how are you?",
         ),
         # Size above STREAM_RESPONSE_CHUNKS
         (

tests/components/tts/test_init.py

@@ -1835,7 +1835,7 @@ async def test_async_convert_audio_error(hass: HomeAssistant) -> None:
     async def bad_data_gen():
         yield bytes(0)
 
-    with pytest.raises(RuntimeError):
+    with pytest.raises(HomeAssistantError):
         # Simulate a bad WAV file
         async for _chunk in tts._async_convert_audio(
             hass, "wav", bad_data_gen(), "mp3"

tests/components/wyoming/snapshots/test_tts.ambr

@@ -1,6 +1,19 @@
 # serializer version: 1
 # name: test_get_tts_audio
   list([
+    dict({
+      'data': dict({
+      }),
+      'payload': None,
+      'type': 'synthesize-start',
+    }),
+    dict({
+      'data': dict({
+        'text': 'Hello world',
+      }),
+      'payload': None,
+      'type': 'synthesize-chunk',
+    }),
     dict({
       'data': dict({
         'text': 'Hello world',
@@ -8,10 +21,29 @@
       'payload': None,
       'type': 'synthesize',
     }),
+    dict({
+      'data': dict({
+      }),
+      'payload': None,
+      'type': 'synthesize-stop',
+    }),
   ])
 # ---
 # name: test_get_tts_audio_different_formats
   list([
+    dict({
+      'data': dict({
+      }),
+      'payload': None,
+      'type': 'synthesize-start',
+    }),
+    dict({
+      'data': dict({
+        'text': 'Hello world',
+      }),
+      'payload': None,
+      'type': 'synthesize-chunk',
+    }),
     dict({
       'data': dict({
         'text': 'Hello world',
@@ -19,10 +51,29 @@
       'payload': None,
       'type': 'synthesize',
     }),
+    dict({
+      'data': dict({
+      }),
+      'payload': None,
+      'type': 'synthesize-stop',
+    }),
   ])
 # ---
 # name: test_get_tts_audio_different_formats.1
   list([
+    dict({
+      'data': dict({
+      }),
+      'payload': None,
+      'type': 'synthesize-start',
+    }),
+    dict({
+      'data': dict({
+        'text': 'Hello world',
+      }),
+      'payload': None,
+      'type': 'synthesize-chunk',
+    }),
     dict({
       'data': dict({
         'text': 'Hello world',
@@ -30,6 +81,12 @@
       'payload': None,
       'type': 'synthesize',
     }),
+    dict({
+      'data': dict({
+      }),
+      'payload': None,
+      'type': 'synthesize-stop',
+    }),
   ])
 # ---
 # name: test_get_tts_audio_streaming
@@ -71,6 +128,23 @@
 # ---
 # name: test_voice_speaker
   list([
+    dict({
+      'data': dict({
+        'voice': dict({
+          'name': 'voice1',
+          'speaker': 'speaker1',
+        }),
+      }),
+      'payload': None,
+      'type': 'synthesize-start',
+    }),
+    dict({
+      'data': dict({
+        'text': 'Hello world',
+      }),
+      'payload': None,
+      'type': 'synthesize-chunk',
+    }),
     dict({
       'data': dict({
         'text': 'Hello world',
@@ -82,5 +156,11 @@
       'payload': None,
       'type': 'synthesize',
     }),
+    dict({
+      'data': dict({
+      }),
+      'payload': None,
+      'type': 'synthesize-stop',
+    }),
   ])
 # ---

tests/components/wyoming/test_tts.py

@@ -52,6 +52,7 @@ async def test_get_tts_audio(
 
     # Verify audio
     audio_events = [
+        AudioStart(rate=16000, width=2, channels=1).event(),
         AudioChunk(audio=audio, rate=16000, width=2, channels=1).event(),
         AudioStop().event(),
     ]
@@ -77,7 +78,10 @@ async def test_get_tts_audio(
         assert wav_file.getframerate() == 16000
         assert wav_file.getsampwidth() == 2
         assert wav_file.getnchannels() == 1
-        assert wav_file.readframes(wav_file.getnframes()) == audio
+
+        # nframes = 0 due to streaming
+        assert len(data) == len(audio) + 44  # WAVE header is 44 bytes
+        assert data[44:] == audio
 
     assert mock_client.written == snapshot
 
@@ -88,6 +92,7 @@ async def test_get_tts_audio_different_formats(
     """Test changing preferred audio format."""
     audio = bytes(16000 * 2 * 1)  # one second
     audio_events = [
+        AudioStart(rate=16000, width=2, channels=1).event(),
         AudioChunk(audio=audio, rate=16000, width=2, channels=1).event(),
         AudioStop().event(),
     ]
@@ -123,6 +128,7 @@ async def test_get_tts_audio_different_formats(
 
     # MP3 is the default
     audio_events = [
+        AudioStart(rate=16000, width=2, channels=1).event(),
         AudioChunk(audio=audio, rate=16000, width=2, channels=1).event(),
         AudioStop().event(),
     ]
@@ -167,6 +173,7 @@ async def test_get_tts_audio_audio_oserror(
     """Test get audio and error raising."""
     audio = bytes(100)
     audio_events = [
+        AudioStart(rate=16000, width=2, channels=1).event(),
         AudioChunk(audio=audio, rate=16000, width=2, channels=1).event(),
         AudioStop().event(),
     ]
@@ -197,6 +204,7 @@ async def test_voice_speaker(
     """Test using a different voice and speaker."""
     audio = bytes(100)
     audio_events = [
+        AudioStart(rate=16000, width=2, channels=1).event(),
         AudioChunk(audio=audio, rate=16000, width=2, channels=1).event(),
         AudioStop().event(),
     ]