Editing transcripts is an amazingly useful feature. A language model should be able to identify bad takes, where the speaker stumbles and/or end mid sentence. When the speaker repeats him/her self it can keep the last take. This would also speed up voice over editing.
AdChoices