Why do you even need a second model for that? There are many simpler ways to slow down the rate that the near-immediate response is pseudo-typed onto the screen.