3 ms·

With next-gen speech-to-text (voicea), it is modestly possible. Am sure that could get better by time!