One upon a time we thought the worst things that could happen during a Zoom conference were accidentally leaving the microphone on while cursing out your cat, hearing someone snoring during your stellar summation of your latest project, or standing up to run to the kitchen while forgetting you have no pants on.
There was previous (german?) research that was able to do this from just well-recorded sound.
HRTF etc. wasn’t required.
https://www.newscientist.com/article/dn7996-keyboard-sounds-reveal-their-words/ (Paywall, apologies, and it’s US, I couldn’t find the german one)