Accelerated transcription, word-level timestamps and diarization with whisperX large-v3 for large audio files. Fork of victor-upmeet/whisperx-a40-large with diarization fixed and group_segments implemented