AFK started with a simple frustration.

Why it exists

I tried Wispr Flow, Monologue, and other apps. Paying every month for something I wanted to use more and more wasn't digestible; it stopped me using the technology. Open-source apps didn't give me the transcription quality, interface, and features I wanted. So I built AFK.

The first version was about one gesture: hold a key, say the sentence, and get back to the thing you were doing. That same speech engine now handles longer recordings and a command-line path, because those are useful jobs too.

AFK is designed and maintained by Keshav.

What AFK does today

• System-wide dictation on macOS, with a configurable hold key and hands-free mode.
• Local transcription for supported audio files, including multi-hour recordings processed in chunks.
• Plain-text, SRT, and JSON exports with word-level timing data when available.
• A CLI that accepts supported audio and MP4 files with audio, sends transcript output to stdout, and works with scripts and coding agents.

It is deliberately a focused tool. The homepage leads with dictation, long recordings, the CLI, and one-time pricing because those are the reasons to use it.

The boundaries

AFK currently targets macOS 15 or later on Apple Silicon. The first run downloads and optimizes the speech model, and the app needs Microphone and Accessibility permissions for its core workflows. Audio processing happens on the Mac after setup, while license activation and update checks are network operations.

The app's file picker is for supported audio inputs. The CLI additionally accepts an MP4 when it contains an audio track; it does not promise arbitrary video formats. For the exact current behavior, start with the transcription guide and CLI reference; for network details, read the privacy page.

Back to AFK home →