Speech to text, a model in the middle, text to speech, fast enough to feel like talking.
Source: OpenAI Whisper (GitHub)