Quentin Fuxa
e61d1d111f
script to lib
2025-03-19 15:10:05 +01:00
Quentin Fuxa
5624c1f6b7
Refactor import statement for AudioProcessor and update cleanup method to be awaited; remove unused formatters and state management files
2025-03-19 11:18:12 +01:00
Quentin Fuxa
7679370cf6
Refactor AudioProcessor methods for improved async handling and WebSocket integration
2025-03-19 10:59:50 +01:00
Quentin Fuxa
5ca65e21b7
Refactor DiartDiarization initialization and streamline WebSocket audio processing
2025-03-19 10:33:22 +01:00
Quentin Fuxa
dc02bcdbdd
refacto 0
2025-03-18 18:31:23 +01:00
Quentin Fuxa
4f87ac3ea4
Refactor PCM conversion to a dedicated function; immediate chunk addition to the diarization queue
2025-03-17 11:46:45 +01:00
Quentin Fuxa
b768b219fe
Warmup functionality: add timeout option (for VM not connected to internet); False option to disable warmup
2025-03-14 11:41:18 +01:00
Silas Kieser
cb5cf39336
fix #84
2025-03-13 15:03:16 +01:00
Quentin Fuxa
3024a9bdb2
Diarization : Uses a rx observer instead of diart attach_hooks method
2025-03-13 12:02:18 +01:00
Quentin Fuxa
e3550ef07d
use confidence scores returned by whisper to immediately validate tokens
2025-03-03 12:08:56 +01:00
Quentin Fuxa
b37d3cafb3
ffmpeg timout from 5 to 15s; diarization lag does not stay = 0 at the beginning
2025-03-03 10:34:32 +01:00
Quentin Fuxa
2a869cd509
solve https://github.com/QuentinFuxa/whisper_streaming_web/issues/60#issuecomment-2692191781
2025-03-01 15:53:35 +01:00
Quentin Fuxa
2d2a4967e6
update import paths
2025-02-28 18:41:12 +01:00
Quentin Fuxa
14af47e84b
undiarized text is assigned to last speaker, with buffer information; traceback is used to format_exc errors
2025-02-28 18:11:36 +01:00
Quentin Fuxa
7b1c88589e
// execution for diarization and transcription
2025-02-28 15:43:46 +01:00
Quentin Fuxa
091d5d7bf5
new buffer format
2025-02-26 21:27:39 +01:00
Quentin Fuxa
2ced4fef20
diarization now works at word - not chunk - level!
2025-02-24 00:35:42 +01:00
Quentin Fuxa
296327071d
Enhance diarization logic to improve speaker attribution : corrects several bugs
2025-02-23 23:14:10 +01:00
Quentin Fuxa
fa39eda923
DiartDiarization loaded in asynccontextmanager lifespan
2025-02-19 16:47:34 +01:00
Quentin Fuxa
2608abf0f3
Improve speaker handling; update sleep duration and manage speaker transitions more effectively
2025-02-19 14:41:37 +01:00
Quentin Fuxa
6121083549
add parameter to disable transcription (only diarization), add time in output
2025-02-19 11:21:40 +01:00
Quentin Fuxa
525abcbca7
Implement logging for WebSocket events and FFmpeg process management; remove obsolete test file
2025-02-15 15:21:33 +01:00
Quentin Fuxa
84b09bb2cc
handle ffmpeg timeouts > 5s
2025-02-14 17:22:03 +00:00
Quentin Fuxa
15089c80fd
Add get_buffer method to retrieve unvalidated buffer in string format
2025-02-12 23:36:52 +00:00
Quentin Fuxa
788fe1c676
optimize ffmpeg buffer reading :round duration to nearest lower 0.1s
2025-02-12 23:28:58 +00:00
Quentin Fuxa
149d2ee44c
Use lifespan to load the model just one
2025-02-12 05:53:55 +01:00
Quentin Fuxa
adaca751ce
remove logs handler
2025-02-12 05:46:56 +01:00
Quentin Fuxa
1f6119e405
all text-related classes now share a common TimedText base class
2025-02-07 23:12:04 +01:00
Quentin Fuxa
f7f1f259c1
buffer is now transcript.text
2025-02-07 12:25:05 +01:00
Silas Kieser
f4d3df3d87
change log format
2025-01-28 21:25:17 +01:00
Silas Kieser
9c7d429e15
add logging config to server
2025-01-28 17:38:13 +01:00
Quentin Fuxa
4622fe7aff
Merge branch 'main' into fix-sentencesegmenter
2025-01-28 15:53:10 +01:00
Silas Kieser
8ee1488c08
rename to_flush to concatenate_tsw
2025-01-27 16:49:22 +01:00
Quentin Fuxa
fba37eba0a
move to src
2025-01-19 21:17:55 +01:00
Quentin Fuxa
5523b51fd7
first speaker is "0" no more None
2025-01-19 19:40:09 +01:00
Quentin Fuxa
977436622a
add diarization (beta). Disabled by default
2025-01-19 17:12:40 +01:00
Quentin Fuxa
ce56264241
split whisper_online.py into smaller files
2025-01-14 20:52:53 +01:00
Quentin Fuxa
9cbac96c44
del online once webstreaming is finished
2025-01-14 20:20:22 +01:00
Ava
f52a5ae3c2
specify encoding to ensure Python reads file as UTF-8
...
executing `python whisper_fastapi_online_server.py --host 0.0.0.0 --port 8000` resulted in error on my setup for me:
```
whisper_streaming_web\whisper_fastapi_online_server.py, line 47, in <module>
html = f.read()
^^^^^^^^
File "C:\Python312\Lib\encodings\cp1252.py", line 23, in decode
return codecs.charmap_decode(input,self.errors,decoding_table)[0]
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
UnicodeDecodeError: 'charmap' codec can't decode byte 0x8f in position 1818: character maps to <undefined>
```
On Windows, Python defaults to the `cp1252` encoding, which may not match the encoding of the file being read.
Files containing special characters, non-ASCII text, or saved with UTF-8 encoding can trigger this error when read without specifying the correct encoding.
2025-01-13 23:12:38 +11:00
Quentin Fuxa
aa0ba598f0
no online conflict when multiple users
2025-01-03 14:48:45 +01:00
silask
5fdb08edae
black formating
2024-12-30 21:53:33 +01:00
Quentin Fuxa
122368bff3
Append full transcription in websocket processing
2024-12-30 15:21:00 +01:00
Quentin Fuxa
c960d1571d
Batch unprocessed audio to reduce Whisper streaming calls
2024-12-28 18:32:27 +01:00
Quentin Fuxa
1cea20a42d
/ws to /asr to distinguish protocol ws:// from endpoint
2024-12-24 19:36:20 +01:00
Quentin Fuxa
bee27c68e6
better buffer gestion
2024-12-19 10:19:24 +01:00
Quentin Fuxa
104f7bde03
add fastapi server with live webm to pcm conversion and web page showing both complete transcription and partial transcription
2024-12-16 13:24:27 +01:00
