Merge pull request #1289 from pipecat-ai/aleix/tts-http-improvements

small TTS http improvements

src/pipecat/services/elevenlabs.py

@@ -570,10 +570,12 @@ class ElevenLabsHttpTTSService(TTSService):
 
                 await self.start_tts_usage_metrics(text)
 
-                yield TTSStartedFrame()
+                # Process the streaming response
+                CHUNK_SIZE = 1024
 
-                async for chunk in response.content:
-                    if chunk:
+                yield TTSStartedFrame()
+                async for chunk in response.content.iter_chunked(CHUNK_SIZE):
+                    if len(chunk) > 0:
                         await self.stop_ttfb_metrics()
                         yield TTSAudioRawFrame(chunk, self.sample_rate, 1)
         except Exception as e:

src/pipecat/services/openai.py

@@ -530,8 +530,10 @@ class OpenAITTSService(TTSService):
 
                 await self.start_tts_usage_metrics(text)
 
+                CHUNK_SIZE = 1024
+
                 yield TTSStartedFrame()
-                async for chunk in r.iter_bytes(8192):
+                async for chunk in r.iter_bytes(CHUNK_SIZE):
                     if len(chunk) > 0:
                         await self.stop_ttfb_metrics()
                         frame = TTSAudioRawFrame(chunk, self.sample_rate, 1)

src/pipecat/services/playht.py

@@ -396,8 +396,6 @@ class PlayHTHttpTTSService(TTSService):
 
         try:
             options = self._create_options()
-            b = bytearray()
-            in_header = True
 
             await self.start_ttfb_metrics()
 
@@ -412,6 +410,8 @@ class PlayHTHttpTTSService(TTSService):
 
             yield TTSStartedFrame()
 
+            b = bytearray()
+            in_header = True
             async for chunk in playht_gen:
                 # skip the RIFF header.
                 if in_header:
@@ -426,11 +426,10 @@ class PlayHTHttpTTSService(TTSService):
                             fh.read(size)
                             (data, size) = struct.unpack("<4sI", fh.read(8))
                         in_header = False
-                else:
-                    if len(chunk):
-                        await self.stop_ttfb_metrics()
-                        frame = TTSAudioRawFrame(chunk, self.sample_rate, 1)
-                        yield frame
+                elif len(chunk) > 0:
+                    await self.stop_ttfb_metrics()
+                    frame = TTSAudioRawFrame(chunk, self.sample_rate, 1)
+                    yield frame
         except Exception as e:
             logger.error(f"{self} error generating TTS: {e}")
         finally:

src/pipecat/services/rime.py

@@ -407,10 +407,10 @@ class RimeHttpTTSService(TTSService):
                 yield TTSStartedFrame()
 
                 # Process the streaming response
-                chunk_size = 8192
+                CHUNK_SIZE = 1024
 
-                async for chunk in response.content.iter_chunked(chunk_size):
-                    if chunk:
+                async for chunk in response.content.iter_chunked(CHUNK_SIZE):
+                    if len(chunk) > 0:
                         await self.stop_ttfb_metrics()
                         frame = TTSAudioRawFrame(chunk, self.sample_rate, 1)
                         yield frame

src/pipecat/services/xtts.py

@@ -150,8 +150,10 @@ class XTTSService(TTSService):
 
             yield TTSStartedFrame()
 
+            CHUNK_SIZE = 1024
+
             buffer = bytearray()
-            async for chunk in r.content.iter_chunked(1024):
+            async for chunk in r.content.iter_chunked(CHUNK_SIZE):
                 if len(chunk) > 0:
                     await self.stop_ttfb_metrics()
                     # Append new chunk to the buffer.

src/pipecat/transports/base_output.py

@@ -232,6 +232,9 @@ class BaseOutputTransport(FrameProcessor):
             await self.push_frame(BotStoppedSpeakingFrame())
             await self.push_frame(BotStoppedSpeakingFrame(), FrameDirection.UPSTREAM)
             self._bot_speaking = False
+            # Clean audio buffer (there could be tiny left overs if not multiple
+            # to our output chunk size).
+            self._audio_buffer = bytearray()
 
     #
     # Sink tasks