hey-jev
Mac voice assistant where one Jev call per utterance maps speech to app, volume, Spotify and system actions; an LLM only answers questions.
Overview
Hey Jev is a voice assistant for macOS. Say "Hey Jev" or hold right Option, speak, and it performs the action and answers back. Jev makes every decision in one call (the README cites $0.00004 per request), local faster-whisper transcribes speech, Fish Audio speaks replies with emotion tags, and an LLM only wakes when Jev decides you asked a question rather than gave a command.
It can open, quit, hide, minimise or switch apps, open browser tabs or websites, control Mac and Spotify volume and playback, toggle dark mode, lock or sleep the Mac, and handle two instructions in one sentence such as pausing Spotify and opening Slack. It also supports timers and reminders and a dictation mode that pastes transcribed text where the cursor is. Apps and mis-heard names are configurable. It controls the Mac through AppleScript and the Keychain, so it is macOS-only.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Which action does this utterance ask for (open app, volume, Spotify, dark mode, lock...)?(inferred) | — |
| Yes / no | Is this a question for the LLM rather than a command? | yes · no |
How it works
- Wake word or hotkey; Whisper transcribes locally
- One Jev call maps the utterance to actions
- Execute via AppleScript; spoken reply via Fish Audio
- Questions are passed to an LLM
Strengths
- Very low cost per request as reported
- Handles two commands in one sentence
- LLM used only for questions
Caveats
- macOS only
- Requires several service keys (TypeSafe, Fish Audio, OpenRouter for dictation)
Getting started
Install Python, create a venv, .venv/bin/pip install -r requirements.txt, add keys, and run per the README.At a glance
- Type
- App
- Status
- prototype
- Language
- Python
- Runtime
- macOS app
- Has code
- yes
Use cases and integrations
How curated lists describe it
Open, quit, hide, minimise or switch to apps, open a new browser tab or a website ("open youtube.com in Brave"), Mac volume up / down / mute / set, Spotify volume, play / pause / next / previous, dark mode, lock or sleep the Mac. Two things in one sentence work too: "pause Spotify and open Slack".
logicrw/awesome-jev-projectsMac voice assistant that maps what you said onto an action - apps, windows, volume, Spotify, dark mode - and handles two instructions in one sentence.
MrJev/awesome-jev
Frequently asked questions
What does Jev decide?
Which action your utterance asks for and whether it is a question for the LLM.
Which platforms?
macOS Sequoia and Tahoe only.
How much does a request cost?
About $0.00004 per Jev request, per the README.
Related projects
Cross-platform dictation app where a decision model chooses whether a personal-dictionary replacement actually fits the sentence.
ProjectbetaPythonLiquid Glass voice typing for macOS 26 with an optional Jev fast path that types clean short sentences without waiting for LLM cleanup.
ProjectbetaSwiftAbleton Live controller where a one-sentence command in English or Japanese is turned into mixer, transport, clip and device actions.
ProjectbetaLocal-first desktop and Android attention workbench that extracts deadlines from WeChat and campus notices; Jev optional.
Projectbeta- jev-canvas★ 16
Draw on a tldraw canvas by voice while pointing a webcam-tracked finger; Jev decides command, action, shape, colour, target and place.
DemoprototypeTypeScript Browser teleprompter that checks off talking points as you cover them, scoring each with a parallel Jev Score question.
DemoprototypeJavaScript