search

SYSTRAN / faster-whisper

Faster Whisper transcription with CTranslate2
MIT License
10.25k stars 859 forks source link

segment /chunk is too long #681

Open sd816224 opened 5 months ago

sd816224 commented 5 months ago

hi there, each segment of srt output is a bit too long for me, then i turn on vad_filter and adjust the max_speech_duration_s to 2. it still output segment much more than 2 seconds.

not sure if the function (get_speech_timestamps) of VadOptions works properly here to split the long speech chunk? thanks a lot

Purfview commented 5 months ago

i turn on vad_filter

VAD filter has nothing to do with text length.

segment of srt output

This repo doesn't have srt writer, you should ask at the repo which srt writer you are using.

You can try Standalone Faster-Whisper from there -> https://github.com/Purfview/whisper-standalone-win Try --standard or --standard_asia, or other settings for srt/vtt output.