I recently made an unusual audio processing challenge where the goal is to recover the English text that a user is typing, just by analysing the recorded sound of their keyboard [0]. Not sure how hard it is. I am able to solve it with my own audio processing tools. If interested - give it a try.
I am very much interested in this. Would you be willing to work with me offline on this? It's an area of research for me that I find fascinating, yet I have difficulty finding information for legitimate reproduction of the techniques that I hear about.
[0] https://ggerganov.github.io/keytap-challenge/