Get the most out of Lispr.

1. Microphone

Lispr needs to hear you. macOS will prompt the first time; if you missed the prompt or revoked the permission, grant it in System Settings → Privacy & Security → Microphone — toggle Lispr on.

2. Accessibility

Lispr inserts the recognized text where your cursor is — that requires the Accessibility permission. Grant it in System Settings → Privacy & Security → Accessibility — toggle Lispr on.

3. Input Monitoring (only on macOS 10.15+)

So Lispr can hear when you hold the trigger key — even when another app has focus. Grant it in System Settings → Privacy & Security → Input Monitoring. Lispr requests this properly through IOHIDRequestAccess since 0.33.

Push to talk

Hold the right ⌥ key, speak, release. Text appears at your cursor in any app. That's it — there's no wake word, no listening mode, no settings.

Latched (hands-free) mode

Double-tap the trigger key and Lispr keeps recording until you tap again. Useful when you need both hands free.

Cancel mid-recording

Press Esc while recording to drop what you've said — nothing gets transcribed, nothing gets inserted.

Alternative trigger keys

If right ⌥ is taken by another shortcut on your Mac, Lispr also supports the Fn key as the push-to-talk trigger.

Languages

Lispr transcribes via Whisper, which is multilingual. The language is detected automatically — you can even switch mid-sentence and it follows. About 99 languages are recognised, no setup required.

Smart sentence joining

If you dictate one phrase, then dictate again right after, Lispr inserts a period and a space between them — no run-on sentences. It only auto-joins after its own previous insertion, so it doesn't touch text you typed yourself.

Clipboard

Lispr uses the clipboard internally to insert text — but it restores whatever you had copied immediately afterwards. Your ⌘V still gives back the link or paragraph you copied an hour ago.

Updates

Lispr auto-updates via Sparkle — when a new release is signed, your installed Lispr offers it the next time the app launches. See the changelog for what changed in each version.

The trigger key doesn't do anything

• Check that Input Monitoring is granted (Step 3 above) — without it Lispr can't see the key press from a background app.
• Make sure no other app is grabbing right ⌥ — Karabiner, BetterTouchTool, Raycast hotkeys, Magnet etc. If so, switch Lispr to Fn from the menu-bar icon.

"We didn't catch that"

Lispr filters very short and silent recordings (since 0.39) — about 9 out of 10 of these were brief silences, not missed words. If you get this and you did speak: hold the key a moment longer before talking, and check that the right input device is selected in the menu-bar icon.

Text ends up in the wrong app

Lispr inserts at whatever cursor is focused when you release the trigger key. If you switched apps mid-recording, the text goes to the new app. Tip: keep your cursor in the destination before holding the key.

The menu-bar icon is missing

• On macOS 11–12 specifically, an earlier build (≤ 0.24) could lose the icon — update to the latest from the changelog page; this was fixed in 0.25.
• If you're on a recent macOS and still don't see it: your menu bar might be full and the icon is hidden under Control Center / Bartender. Try widening the menu bar by hiding a few icons.

Transcription timeout

If your network blips, Lispr now bails out after 10 seconds (since 0.34) and shows a friendly message. Try again — the next request often succeeds because Lispr also auto-retries on transient failures (since 0.31).

Sparkle can't fetch the update

The DMG and appcast are served from proxy.lispr.ai over a TCP-only path (since 0.39) — this bypasses HTTP/3 issues some networks have. If your network still blocks it, you can always download the latest DMG directly from lispr.ai and replace the app manually.

The 10-minute recording cap

Lispr stops a single recording after 10 minutes with a localized warning (since 0.37) — that's a soft cap to keep accidental marathon recordings from going through. Release and start a new recording for longer dictations.

Where does my voice go?

From your Mac it goes over an encrypted connection to an edge proxy we run, which forwards it to Groq for transcription. Groq returns the text, the proxy returns it to your Mac, and the audio is discarded immediately. Nothing is stored on a server. Nothing trains a model.

Is there an account?

No. No sign-up, no password, no profile. There's nothing to log into because there's nothing to log into.

What about logs?

Since 0.40, Lispr can send diagnostic X-Lispr-* headers when something goes wrong — these are headers (timing, error codes), not audio or transcripts, and we don't persist them server-side.

Does it work offline?

No — transcription runs in the cloud, which is why it's both fast (~300 ms median end-to-end since 0.41) and accurate. If you need fully-local dictation, macOS's built-in Dictation is a fine alternative; that's the trade-off we made for speed and language coverage.