search

NavodPeiris / speechlib

speechlib is a library that can do speaker diarization, transcription and speaker recognition on an audio file to create transcripts with actual speaker names
MIT License
114 stars 10 forks source link

faster-whisper==1.0.1 transcription creates duplicate sentences. #32

Closed AnshumanParidaIL closed 1 month ago

AnshumanParidaIL commented 1 month ago

Transcribing with faster-whisper==1.0.1 produces duplicate sentences: image

This issue is solved by upgrading faster-whisper to 1.0.2 as discussed here in the official SYSTRAN/faster-whisper repo.

NavodPeiris commented 1 month ago

in latests speechlib==1.0.15 version, this is fixed and now using faster-whisper==1.0.2. We were using faster-whisper==0.10.1

Thank you for the heads up!!

AnshumanParidaIL commented 1 month ago

Thank you for the clarification.