This is a modified version of Runpod's Faster Whisper worker to change the format to return only Craig needs, put word timestamps in segments, and allow for multiple audio inputs.
This also incorporates a modified method of assembling (or correcting) multiple transcriptions into a single list of lines from Kadda OK's TASMAS. This uses Deep Multilingual Punctuation Prediction using the kredor/punctuate-all model for correcting punctuations.