audiOS

The keyboard replaces letter keys with three controls modeled on a cassette deck. REC starts and stops a take. The unsay rocker rolls back the last chunk of text. The log wheel sends or saves your words. The result is a dictation setup you can drive entirely by feel, without ever looking down to hunt for a tiny on-screen icon during a conversation.

Words appear as you speak rather than after you stop. You keep your train of thought going, and you catch a bad transcription the moment it shows up. audioS runs the audio through Groq's Whisper API, which handles the heavy speech-to-text work behind the scenes.

Mistakes happen when you talk. Flick the rocker and audioS removes the last bit you said, letting you re-say it without deleting character by character. It's a small thing that makes speaking feel less final than typing. That's the whole point.

The log wheel doesn't just send. It keeps a record of your takes, so a captured thought stays available even after the message goes out. That's handy when you want to reuse a phrase later or double-check what you actually said before someone responds to it.

Because audioS is a system keyboard, it shows up anywhere you can type: Messages, Mail, Notes, social apps, the whole list. You switch to it like any other keyboard, and there's no copying between apps or jumping into a separate dictation screen every time you want to speak.

audioS collects anonymous usage signals such as button-press counts, audio duration, and response times. It states plainly that no keystroke content or transcription text is stored. The vendor ID it uses to enforce rate limits isn't tied to your Apple ID.