public-mirrors/Kokoro-FastAPI
1
0
Fork 0
mirror of https://github.com/remsky/Kokoro-FastAPI.git synced 2025-08-05 16:48:53 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Fix bugs of generating empty file when using streaming mode.

Browse source 
The reason is that when stream=True, the audio conversion functions are not really called.
This commit is contained in:
CodePothunter 2025-03-09 14:12:18 +08:00
parent 3e6ee65482
commit f998cf8d01
2 changed files with 22 additions and 6 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

19
api/src/routers/openai_compatible.py
View file

@ -166,7 +166,6 @@ async def stream_audio_chunks(
normalization_options=request.normalization_options, normalization_options=request.normalization_options,
return_timestamps=unique_properties["return_timestamps"], return_timestamps=unique_properties["return_timestamps"],
): ):
# Check if client is still connected # Check if client is still connected
is_disconnected = client_request.is_disconnected is_disconnected = client_request.is_disconnected
if callable(is_disconnected): if callable(is_disconnected):
@ -174,7 +173,7 @@ async def stream_audio_chunks(
if is_disconnected: if is_disconnected:
logger.info("Client disconnected, stopping audio generation") logger.info("Client disconnected, stopping audio generation")
break break
yield chunk_data yield chunk_data
except Exception as e: except Exception as e:
logger.error(f"Error in audio streaming: {str(e)}") logger.error(f"Error in audio streaming: {str(e)}")
@ -273,9 +272,21 @@ async def create_speech(
async def single_output(): async def single_output():
try: try:
# Stream chunks # Stream chunks
is_first_chunk=True
async for chunk_data in generator: async for chunk_data in generator:
if chunk_data.output: # Skip empty chunks if chunk_data.audio is not None and len(chunk_data.audio) > 0: # Skip empty chunks
yield chunk_data.output # Convert to requested format with proper encoding
encoded_chunk = await AudioService.convert_audio(
chunk_data,
24000,
request.response_format,
is_first_chunk=is_first_chunk,
is_last_chunk=False,
trim_audio=False
)
if encoded_chunk.output:
yield encoded_chunk.output
is_first_chunk=False
except Exception as e: except Exception as e:
logger.error(f"Error in single output streaming: {e}") logger.error(f"Error in single output streaming: {e}")
raise raise

9
api/src/services/streaming_audio_writer.py
View file

@ -64,9 +64,14 @@ class StreamingAudioWriter:
frame.pts = self.pts frame.pts = self.pts
self.pts += frame.samples self.pts += frame.samples
encoded_data = b""
for packet in self.stream.encode(frame): for packet in self.stream.encode(frame):
self.container.mux(packet) self.container.mux(packet)
# Get the encoded data from the buffer
encoded_data = self.output_buffer.getvalue()
# Clear the buffer for next write
self.output_buffer.seek(0)
self.output_buffer.truncate(0)
# Only return a null byte and keep the container running. return encoded_data
return b""