A macOS menubar app for local speech-to-text. Hold a hotkey, speak, release — your transcription is pasted into the focused app.
- Fully local transcription via WhisperKit (CoreML, Apple Silicon)
- Custom dictionary for domain-specific terms
- Runs in the menu bar, starts at login
Everything runs on-device. Audio is transcribed locally with WhisperKit and never leaves your Mac. The only network request the app makes is the one-time WhisperKit model download on first launch (or you can bundle the model so there's no download at all).
- macOS 14+ (Sonoma)
- Apple Silicon (M1/M2/M3/M4)
- Xcode Command Line Tools (
xcode-select --install)
- Download the latest
ainstype-x.y.z.dmgfrom the Releases page. - Open the DMG and drag ainstype into your Applications folder.
- Launch ainstype from Applications — it runs in the menu bar (no dock icon).
- Grant Input Monitoring and Accessibility permissions when prompted.
Hold Right Cmd and speak. Release to transcribe and paste.
The DMG bundles the WhisperKit model, so there's no first-run download.
swift build && swift runOn first launch, the app downloads the WhisperKit model (~616MB) unless one is already bundled.
./build_app.shThis builds a signed .app bundle and creates a DMG. Optionally bundles the WhisperKit model so users skip the first-run download.
Edit ~/.config/ainstype/config.toml:
language = "en" # optional, auto-detect if omitted
[recording]
hotkey = "cmd_r" # cmd, cmd_r, alt, alt_r, ctrl, ctrl_rCustom terms in ~/.config/ainstype/dictionary.toml:
[terms]
words = ["Kubernetes", "TypeScript", "PostgreSQL"]
[replacements]
"kube control" = "kubectl"
"post gress" = "Postgres"