What i'm requesting is a special mode of AI transcription that ONLY fills in the gaps. (after a human transcript has been sync'd).

E.g, it only identifies audio sections that aren't sync'd

Visually these are the black sections you will see when the video starts to play (see images attached).

Or see this screen cast...

---

Discussion

I add a human transcript but it doesn't HAVE the non-word fillers like ummms.

It TYPICALLY is also missing other fillers like repeats "you you know and and that's why we we... "

It's clear from the timeline that Descript "knows" where in the time code that where there are sections of audio that didn't sync'd to the human transcript (see picture).

The request would be an option for an AI pass after Sync'ing an external "human" transcription.

The second pass AI transcription would limit itself to the un-sync'd sections.