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 streaming a wav file with captions not reaturning any captions (This is only a problem because wav streaming does not acually work)

Browse source
This commit is contained in:
Fireblade 2025-02-16 16:49:33 -05:00
parent e3dc959775
commit cb22aab239
1 changed files with 22 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

21
api/src/routers/development.py
View file

@ -210,11 +210,21 @@ async def create_captioned_speech(
try: try:
# Write chunks to temp file and stream # Write chunks to temp file and stream
async for chunk_data in generator: async for chunk_data in generator:
# The timestamp acumulator is only used when word level time stamps are generated but no audio is returned.
timestamp_acumulator=[]
if chunk_data.output: # Skip empty chunks if chunk_data.output: # Skip empty chunks
await temp_writer.write(chunk_data.output) await temp_writer.write(chunk_data.output)
base64_chunk= base64.b64encode(chunk_data.output).decode("utf-8") base64_chunk= base64.b64encode(chunk_data.output).decode("utf-8")
# Add any chunks that may be in the acumulator into the return word_timestamps
chunk_data.word_timestamps=timestamp_acumulator + chunk_data.word_timestamps
timestamp_acumulator=[]
yield CaptionedSpeechResponse(audio=base64_chunk,audio_format=content_type,timestamps=chunk_data.word_timestamps) yield CaptionedSpeechResponse(audio=base64_chunk,audio_format=content_type,timestamps=chunk_data.word_timestamps)
else:
if chunk_data.word_timestamps is not None and len(chunk_data.word_timestamps) > 0:
timestamp_acumulator+=chunk_data.word_timestamps
# Finalize the temp file # Finalize the temp file
await temp_writer.finalize() await temp_writer.finalize()
@ -234,13 +244,24 @@ async def create_captioned_speech(
async def single_output(): async def single_output():
try: try:
# The timestamp acumulator is only used when word level time stamps are generated but no audio is returned.
timestamp_acumulator=[]
# Stream chunks # Stream chunks
async for chunk_data in generator: async for chunk_data in generator:
if chunk_data.output: # Skip empty chunks if chunk_data.output: # Skip empty chunks
# Encode the chunk bytes into base 64 # Encode the chunk bytes into base 64
base64_chunk= base64.b64encode(chunk_data.output).decode("utf-8") base64_chunk= base64.b64encode(chunk_data.output).decode("utf-8")
# Add any chunks that may be in the acumulator into the return word_timestamps
chunk_data.word_timestamps=timestamp_acumulator + chunk_data.word_timestamps
timestamp_acumulator=[]
yield CaptionedSpeechResponse(audio=base64_chunk,audio_format=content_type,timestamps=chunk_data.word_timestamps) yield CaptionedSpeechResponse(audio=base64_chunk,audio_format=content_type,timestamps=chunk_data.word_timestamps)
else:
if chunk_data.word_timestamps is not None and len(chunk_data.word_timestamps) > 0:
timestamp_acumulator+=chunk_data.word_timestamps
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