It is more like ai model problem(then app logic. doing it more frequently will require more computation. Things like speculative decoding can help though).

Doing it locally is hard, but we expect to ship it very soon. Please join our Discord(https://hyprnote.com/discord) if you are interested to hear from us.