Flip the script, ask the students to record videos instead and then use AI to extract text out of the videos.
Having read the reddit thread linked in the article, I think the main problems would be one of (depending on the teacher):
1. Not wanting to
2. Needing funding/support to AI-transcribe all the video for each of their numerous students, but their institution not providing it.
But those seem [eventually] surmountable. I really like the idea.