100% this. Even better if this could happen inside a mobile app, maybe not fully featured for editing, but at least for interview recording & scripted recording. (I can't record on my laptop because the fan is so damn loud, especially the second you open Descript...)

This this this!

This is how i would see the workflow.

User types plain text into Descript

Select text and hit record

User records audio, Descript uses the text as transcription, flags/highlights if different.

Any other tracks pinned to the text section of the timeline carries over.

Yes, this!

With the new live transcript it could show me what I was planning to say (the pre-written script) and what I actually said (the live transcript) and flag it when I say something different. Then if I re-read the part that is wrong it can automatically correct it. And it could also automatically remove filler words and sounds.

When this is combined with Overdub it could automatically correct mistakes too. This would still be useful when recording video where you are looking into a teleprompter. And the automatic feedback from the live transcription would let me know when I need to fix it without having to do an additional take.