Current product behavior
AFK documentation.
Install the app, get the model ready, choose how dictation behaves, transcribe long audio, and connect the same engine to a shell.
Choose a starting point.
These guides describe the current AFK app and CLI. When a boundary matters, the docs say so.
Getting started
Requirements, signed install, first-run model setup, license activation, permissions, and your first dictation.
System-wide dictation
Hold-to-talk, hands-free sequences, clipboard fallback, cleanup, language, and the custom dictionary.
Notes and history
Search dictations, expand long notes, use visible copy/delete actions, and reopen saved transcription files.
Transcribe audio
Supported inputs, chunking for long recordings, queue controls, result expiry, history references, and exports.
CLI reference
Transcribe audio or MP4, understand output streams, and install the bundled Agent Skill for coding agents.
Troubleshooting
Work through model setup, permission, shortcut, audio-input, missing-result, and CLI problems.
Privacy and network boundaries
See the model download, Polar activation, Sparkle update checks, local history, and one-hour cache behavior.
Quick orientation
AFK is a Mac app first.
It needs macOS 15 or later on Apple Silicon. The speech model is downloaded once, then recognition runs locally. The app's global dictation path needs Microphone and Accessibility permissions.