3 ms·

you would need to tie microphone input with the actual keys typed, and enough of it to train a model. nothingburger