admin-valentin/WhisperLiveKit
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

new buffer format

Browse source
This commit is contained in:
Quentin Fuxa 2025-02-26 21:27:39 +01:00
parent 9d5d6d8031
commit 091d5d7bf5
2 changed files with 5 additions and 3 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
src/whisper_streaming/online_asr.py
View file

@ -168,7 +168,7 @@ class OnlineASRProcessor:
""" """
Get the unvalidated buffer in string format. Get the unvalidated buffer in string format.
""" """
return self.concatenate_tokens(self.transcript_buffer.buffer).text return self.concatenate_tokens(self.transcript_buffer.buffer)
def process_iter(self) -> Transcript: def process_iter(self) -> Transcript:

6
whisper_fastapi_online_server.py
View file

@ -186,7 +186,6 @@ async def websocket_endpoint(websocket: WebSocket):
if not chunk: if not chunk:
logger.info("FFmpeg stdout closed.") logger.info("FFmpeg stdout closed.")
break break
pcm_buffer.extend(chunk) pcm_buffer.extend(chunk)
if len(pcm_buffer) >= BYTES_PER_SEC: if len(pcm_buffer) >= BYTES_PER_SEC:
if len(pcm_buffer) > MAX_BYTES_PER_SEC: if len(pcm_buffer) > MAX_BYTES_PER_SEC:
@ -207,7 +206,9 @@ async def websocket_endpoint(websocket: WebSocket):
new_tokens = online.process_iter() new_tokens = online.process_iter()
tokens.extend(new_tokens) tokens.extend(new_tokens)
full_transcription += sep.join([t.text for t in new_tokens]) full_transcription += sep.join([t.text for t in new_tokens])
buffer = online.get_buffer() _buffer = online.get_buffer()
buffer = _buffer.text
end_buffer = _buffer.end if _buffer.end else tokens[-1].end if tokens else 0
if buffer in full_transcription: # With VAC, the buffer is not updated until the next chunk is processed if buffer in full_transcription: # With VAC, the buffer is not updated until the next chunk is processed
buffer = "" buffer = ""
else: else:
@ -253,6 +254,7 @@ async def websocket_endpoint(websocket: WebSocket):
lines[-1]["diff"] = round(token.end - last_end_diarized, 2) lines[-1]["diff"] = round(token.end - last_end_diarized, 2)
response = {"lines": lines, "buffer": buffer} response = {"lines": lines, "buffer": buffer}
# response = {"lines": lines, "buffer": buffer, "time_buffer_transcription": time() + beg_loop - end_buffer, "time_buffer_diarization": time() + beg_loop - end_attributed_speaker}
await websocket.send_json(response) await websocket.send_json(response)
except Exception as e: except Exception as e: