Jot

What does Jot do?

Jot's README describes it as "Smart dictation for macOS that puts polished text wherever your cursor is". You hold the fn key, say the thing, release, and the words appear in the app you were already using, punctuated and with filler words removed. The README says there is no window to switch to, no transcript to copy and no account to make.

The project is explicitly a Gemini 3.5 Transcribe demo. The README says this is "not an officially supported Google product". It is a menu bar app, a pill at the bottom of the screen while you talk, and a History window. The README highlights that Jot "follows a change of mind": say "let's meet at 1pm, actually no, make it 2pm" and Jot writes "Let's meet at 2pm."

Key facts

• Licence: Apache-2.0 on the code. Bundled fonts Google Sans Flex and Google Sans Code are SIL OFL 1.1. Earcons are Apache-2.0 originals.
• Model: gemini-3.5-transcribe, Google's specialist transcription model. The README says your key needs access to it, and setup tells you up front if it does not.
• Cost: you pay Google for what you dictate at the Gemini API's own pricing. The README says a free tier exists and a typical dictation is a few seconds of audio. Jot itself is free and has no account.
• Platform: macOS 14 or later. Development needs Xcode 16 and up and the xcodegen tool.
• Gestures: hold fn to record, fn plus tap Space for hands-free, Esc to cancel. Anything over 10 seconds is still kept in History. The key is rebindable.
• Crash safety: audio goes to disk from the first millisecond, written as CAF, so a crash, a kill -9 or a flat battery costs nothing. The recording is recovered on next launch.
• Offline: dictations queue and land when you reconnect. Every failure is retryable from History.
• Privacy: your voice goes from your Mac straight to the Gemini API with your own key. One network host; the README says every line of the code that talks to it can be read. API keys are stored in the macOS Keychain.
• Insertion ladder: Accessibility API first, then a guarded paste that restores the clipboard, then a "copied, press Command V" chip. The README says Jot never blind-pastes into an app that stole focus mid-flight.
• Dictionary: names and product terms ride along with the audio, so the model hears "Kubernetes" rather than guessing "cooper netties".
• Transcript text is logged as private and never appears in system logs.

What does it replace, and where does it fall short?

Jot replaces Wispr Flow and Superwhisper for dictating into any Mac app, with polished output and no account. See the open-source alternatives to Wispr Flow page.

Where it falls short: it is not offline transcription. The audio is sent to the Gemini API with your key, so the model runs on Google's servers rather than on your Mac. If you want fully local dictation, BetterWispr, Handy or VoiceInk fit better. Jot is macOS only. The README calls the project a demo, so expect the shape of a demo rather than a fully supported product.

How does Jot run?

Download the latest Jot-x.y.z.dmg from GitHub Releases, drag Jot into Applications, and launch it. Setup takes about two minutes: paste a Gemini API key from Google AI Studio, allow the microphone and Accessibility, and start dictating. Keys are stored in the macOS Keychain and only ever sent to Google.

The maintainers do not run a hosted version. To build from source, install xcodegen with Homebrew and run ./scripts/build.sh.

Who is Jot for?

A Mac user who wants Gemini-quality polished dictation into any app, is comfortable paying Google for API tokens, and does not need offline operation. Someone who wants transcription that never leaves their Mac should look at BetterWispr, Handy or VoiceInk on the open-source alternatives to Wispr Flow page.

What limits does the README state?

Apps run from a mounted disk image are sandboxed by macOS and the permissions granted will not stick, so the README says to install into Applications before launching. Jot requires an active Gemini API key with access to gemini-3.5-transcribe; setup checks this up front rather than failing on the first dictation. The project is marked as a demo and the README states it is not an officially supported Google product.

Questions people ask

Is Jot open source?

Yes, under Apache-2.0, which is an OSI-approved licence. Bundled Google Sans Flex and Google Sans Code fonts are SIL OFL 1.1, and the earcons are original Apache-2.0 works. The README states the licence and the font and asset grants.

Does Jot work offline?

Partly. Recording works offline because audio is written to disk from the first millisecond. Transcription needs a round trip to the Gemini API, so the words only appear once the Mac is online. The README says dictations queue and land when you reconnect.

What does Jot cost to use?

Jot itself is free and has no account. The transcription uses the Gemini API with your key, so you pay Google at the Gemini API's own pricing. The README says a free tier exists and a typical dictation is a few seconds of audio.

How does Jot compare with Wispr Flow?

Wispr Flow is a paid product with its own cloud. Jot is a Mac app that sends audio to Google's Gemini API using your own key. Both let you hold a key and paste polished text into any app; Jot keeps you on your own Google key and Google's pricing, with the code open to read. See the open-source alternatives to Wispr Flow page.

How does Jot keep words safe if something crashes?

The README says audio is written to disk as CAF from the first millisecond and the recording is recovered on next launch. Anything over 10 seconds is kept in History even if cancelled. Offline, dictations queue and land when you reconnect, and every failure can be retried from History.

Sources

• Jot README and LICENSE: github.com/google-gemini/jot-gemini-transcribe-macOS, read 2026-10-12. The README states Apache-2.0, calls the project a Gemini 3.5 Transcribe demo and says it is not an officially supported Google product.