Speech to text, text to speech, screen understanding, on-device assistants, post-ASR dictation cleanup, and evaluation.