services: async generators should yield frames

src/pipecat/services/aws.py

@@ -190,7 +190,7 @@ class AWSTTSService(TTSService):
 
             await self.start_tts_usage_metrics(text)
 
-            await self.push_frame(TTSStartedFrame())
+            yield TTSStartedFrame()
 
             if "AudioStream" in response:
                 with response["AudioStream"] as stream:
@@ -203,7 +203,7 @@ class AWSTTSService(TTSService):
                             frame = TTSAudioRawFrame(chunk, self._settings["sample_rate"], 1)
                             yield frame
 
-            await self.push_frame(TTSStoppedFrame())
+            yield TTSStoppedFrame()
 
         except (BotoCoreError, ClientError) as error:
             logger.exception(f"{self} error generating TTS: {error}")
@@ -211,4 +211,4 @@ class AWSTTSService(TTSService):
             yield ErrorFrame(error=error_message)
 
         finally:
-            await self.push_frame(TTSStoppedFrame())
+            yield TTSStoppedFrame()

src/pipecat/services/azure.py

@@ -261,14 +261,14 @@ class AzureTTSService(TTSService):
         if result.reason == ResultReason.SynthesizingAudioCompleted:
             await self.start_tts_usage_metrics(text)
             await self.stop_ttfb_metrics()
-            await self.push_frame(TTSStartedFrame())
+            yield TTSStartedFrame()
             # Azure always sends a 44-byte header. Strip it off.
             yield TTSAudioRawFrame(
                 audio=result.audio_data[44:],
                 sample_rate=self._settings["sample_rate"],
                 num_channels=1,
             )
-            await self.push_frame(TTSStoppedFrame())
+            yield TTSStoppedFrame()
         elif result.reason == ResultReason.Canceled:
             cancellation_details = result.cancellation_details
             logger.warning(f"Speech synthesis canceled: {cancellation_details.reason}")

src/pipecat/services/cartesia.py

@@ -255,8 +255,8 @@ class CartesiaTTSService(WordTTSService):
                 await self._connect()
 
             if not self._context_id:
-                await self.push_frame(TTSStartedFrame())
                 await self.start_ttfb_metrics()
+                yield TTSStartedFrame()
                 self._context_id = str(uuid.uuid4())
 
             msg = self._build_msg(text=text)
@@ -266,7 +266,7 @@ class CartesiaTTSService(WordTTSService):
                 await self.start_tts_usage_metrics(text)
             except Exception as e:
                 logger.error(f"{self} error sending message: {e}")
-                await self.push_frame(TTSStoppedFrame())
+                yield TTSStoppedFrame()
                 await self._disconnect()
                 await self._connect()
                 return
@@ -331,8 +331,8 @@ class CartesiaHttpTTSService(TTSService):
     async def run_tts(self, text: str) -> AsyncGenerator[Frame, None]:
         logger.debug(f"Generating TTS: [{text}]")
 
-        await self.push_frame(TTSStartedFrame())
         await self.start_ttfb_metrics()
+        yield TTSStartedFrame()
 
         try:
             voice_controls = None
@@ -365,4 +365,4 @@ class CartesiaHttpTTSService(TTSService):
             logger.error(f"{self} exception: {e}")
 
         await self.start_tts_usage_metrics(text)
-        await self.push_frame(TTSStoppedFrame())
+        yield TTSStoppedFrame()

src/pipecat/services/deepgram.py

@@ -84,7 +84,7 @@ class DeepgramTTSService(TTSService):
             )
 
             await self.start_tts_usage_metrics(text)
-            await self.push_frame(TTSStartedFrame())
+            yield TTSStartedFrame()
 
             # The response.stream_memory is already a BytesIO object
             audio_buffer = response.stream_memory
@@ -105,7 +105,7 @@ class DeepgramTTSService(TTSService):
                 )
                 yield frame
 
-            await self.push_frame(TTSStoppedFrame())
+                yield TTSStoppedFrame()
 
         except Exception as e:
             logger.exception(f"{self} exception: {e}")

src/pipecat/services/elevenlabs.py

@@ -387,8 +387,8 @@ class ElevenLabsTTSService(WordTTSService):
 
             try:
                 if not self._started:
-                    await self.push_frame(TTSStartedFrame())
                     await self.start_ttfb_metrics()
+                    yield TTSStartedFrame()
                     self._started = True
                     self._cumulative_time = 0
 
@@ -396,7 +396,7 @@ class ElevenLabsTTSService(WordTTSService):
                 await self.start_tts_usage_metrics(text)
             except Exception as e:
                 logger.error(f"{self} error sending message: {e}")
-                await self.push_frame(TTSStoppedFrame())
+                yield TTSStoppedFrame()
                 await self._disconnect()
                 await self._connect()
                 return

src/pipecat/services/google.py

@@ -361,7 +361,7 @@ class GoogleTTSService(TTSService):
 
             await self.start_tts_usage_metrics(text)
 
-            await self.push_frame(TTSStartedFrame())
+            yield TTSStartedFrame()
 
             # Skip the first 44 bytes to remove the WAV header
             audio_content = response.audio_content[44:]
@@ -377,11 +377,11 @@ class GoogleTTSService(TTSService):
                 yield frame
                 await asyncio.sleep(0)  # Allow other tasks to run
 
-            await self.push_frame(TTSStoppedFrame())
+            yield TTSStoppedFrame()
 
         except Exception as e:
             logger.exception(f"{self} error generating TTS: {e}")
             error_message = f"TTS generation error: {str(e)}"
             yield ErrorFrame(error=error_message)
         finally:
-            await self.push_frame(TTSStoppedFrame())
+            yield TTSStoppedFrame()

src/pipecat/services/lmnt.py

@@ -176,8 +176,8 @@ class LmntTTSService(TTSService):
                 await self._connect()
 
             if not self._started:
-                await self.push_frame(TTSStartedFrame())
                 await self.start_ttfb_metrics()
+                yield TTSStartedFrame()
                 self._started = True
 
             try:
@@ -186,7 +186,7 @@ class LmntTTSService(TTSService):
                 await self.start_tts_usage_metrics(text)
             except Exception as e:
                 logger.error(f"{self} error sending message: {e}")
-                await self.push_frame(TTSStoppedFrame())
+                yield TTSStoppedFrame()
                 await self._disconnect()
                 await self._connect()
                 return

src/pipecat/services/openai.py

@@ -426,13 +426,13 @@ class OpenAITTSService(TTSService):
 
                 await self.start_tts_usage_metrics(text)
 
-                await self.push_frame(TTSStartedFrame())
+                yield TTSStartedFrame()
                 async for chunk in r.iter_bytes(8192):
                     if len(chunk) > 0:
                         await self.stop_ttfb_metrics()
                         frame = TTSAudioRawFrame(chunk, self._settings["sample_rate"], 1)
                         yield frame
-                await self.push_frame(TTSStoppedFrame())
+                yield TTSStoppedFrame()
         except BadRequestError as e:
             logger.exception(f"{self} error generating TTS: {e}")
 

src/pipecat/services/playht.py

@@ -75,7 +75,7 @@ class PlayHTTTSService(TTSService):
 
             await self.start_tts_usage_metrics(text)
 
-            await self.push_frame(TTSStartedFrame())
+            yield TTSStartedFrame()
             async for chunk in playht_gen:
                 # skip the RIFF header.
                 if in_header:
@@ -95,6 +95,6 @@ class PlayHTTTSService(TTSService):
                         await self.stop_ttfb_metrics()
                         frame = TTSAudioRawFrame(chunk, self._settings["sample_rate"], 1)
                         yield frame
-            await self.push_frame(TTSStoppedFrame())
+            yield TTSStoppedFrame()
         except Exception as e:
             logger.exception(f"{self} error generating TTS: {e}")

src/pipecat/services/xtts.py

@@ -151,7 +151,7 @@ class XTTSService(TTSService):
 
             await self.start_tts_usage_metrics(text)
 
-            await self.push_frame(TTSStartedFrame())
+            yield TTSStartedFrame()
 
             buffer = bytearray()
             async for chunk in r.content.iter_chunked(1024):
@@ -187,4 +187,4 @@ class XTTSService(TTSService):
                 frame = TTSAudioRawFrame(resampled_audio_bytes, 16000, 1)
                 yield frame
 
-            await self.push_frame(TTSStoppedFrame())
+            yield TTSStoppedFrame()