Magnus is built to be spoken to. Click the microphone in the composer and talk; your words appear as text and send like any typed message. On Chrome and Edge this uses the browser's native speech recognition; on Safari and Firefox, Magnus falls back to an on-device Whisper model - push to talk, release to transcribe - so voice input works everywhere.
Turn on hands-free in the voice settings and the mic re-arms itself after every answer, so you can hold a whole conversation without touching the keyboard. Combined with voice output you get a genuine back-and-forth: you speak, Magnus answers aloud, and the mic is already listening for your reply.
You can barge in while Magnus is speaking: say two or more words and he stops talking to listen. An echo filter keeps him from hearing himself through your speakers - anything that overlaps heavily with his own last sentence is ignored, while short answers like "yes" or "no" always get through. Headphones make this perfect, but it works fine on laptop speakers too.