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

Fixes a couple of issues with audio triming and prevents errors with single voice weights

Browse source
This commit is contained in:
Fireblade 2025-02-18 18:12:49 -05:00
parent f2b2f41412
commit 4ee4d36822
4 changed files with 11 additions and 3 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

1
api/src/routers/development.py
View file

@ -294,6 +294,7 @@ async def create_captioned_speech(
request.response_format, request.response_format,
is_first_chunk=True, is_first_chunk=True,
is_last_chunk=False, is_last_chunk=False,
trim_audio=False,
) )
# Convert to requested format with proper finalization # Convert to requested format with proper finalization

1
api/src/routers/openai_compatible.py
View file

@ -294,6 +294,7 @@ async def create_speech(
request.response_format, request.response_format,
is_first_chunk=True, is_first_chunk=True,
is_last_chunk=False, is_last_chunk=False,
trim_audio=False
) )
# Convert to requested format with proper finalization # Convert to requested format with proper finalization

3
api/src/services/audio.py
View file

@ -119,6 +119,7 @@ class AudioService:
chunk_text: str = "", chunk_text: str = "",
is_first_chunk: bool = True, is_first_chunk: bool = True,
is_last_chunk: bool = False, is_last_chunk: bool = False,
trim_audio: bool = True,
normalizer: AudioNormalizer = None, normalizer: AudioNormalizer = None,
) -> Tuple[AudioChunk]: ) -> Tuple[AudioChunk]:
"""Convert audio data to specified format with streaming support """Convert audio data to specified format with streaming support
@ -147,6 +148,8 @@ class AudioService:
normalizer = AudioNormalizer() normalizer = AudioNormalizer()
audio_chunk.audio = normalizer.normalize(audio_chunk.audio) audio_chunk.audio = normalizer.normalize(audio_chunk.audio)
if trim_audio == True:
audio_chunk = AudioService.trim_audio(audio_chunk,chunk_text,speed,is_last_chunk,normalizer) audio_chunk = AudioService.trim_audio(audio_chunk,chunk_text,speed,is_last_chunk,normalizer)
# Get or create format-specific writer # Get or create format-specific writer

7
api/src/services/tts_service.py
View file

@ -65,7 +65,7 @@ class TTSService:
yield AudioChunk(np.array([], dtype=np.int16),output=b'') yield AudioChunk(np.array([], dtype=np.int16),output=b'')
return return
chunk_data = await AudioService.convert_audio( chunk_data = await AudioService.convert_audio(
AudioChunk(np.array([0], dtype=np.float32)), # Dummy data for type checking AudioChunk(np.array([], dtype=np.float32)), # Dummy data for type checking
24000, 24000,
output_format, output_format,
speed, speed,
@ -225,6 +225,8 @@ class TTSService:
return voice, combined_path return voice, combined_path
else: else:
# Single voice # Single voice
if "(" in voice and ")" in voice:
voice = voice.split("(")[0].strip()
path = await self._voice_manager.get_voice_path(voice) path = await self._voice_manager.get_voice_path(voice)
if not path: if not path:
raise RuntimeError(f"Voice not found: {voice}") raise RuntimeError(f"Voice not found: {voice}")
@ -341,9 +343,10 @@ class TTSService:
try: try:
async for audio_stream_data in self.generate_audio_stream(text,voice,speed=speed,normalization_options=normalization_options,return_timestamps=return_timestamps,lang_code=lang_code,output_format=None): async for audio_stream_data in self.generate_audio_stream(text,voice,speed=speed,normalization_options=normalization_options,return_timestamps=return_timestamps,lang_code=lang_code,output_format=None):
if len(audio_stream_data.audio) > 0:
audio_data_chunks.append(audio_stream_data) audio_data_chunks.append(audio_stream_data)
combined_audio_data=AudioChunk.combine(audio_data_chunks) combined_audio_data=AudioChunk.combine(audio_data_chunks)
return combined_audio_data return combined_audio_data
except Exception as e: except Exception as e: