Your terminal, now speaking.
ShellVoice is a voice layer for your Mac. Hold a key and dictate into any app. Have your tools speak back. Everything runs on your machine.
Local transcription · Free neural voices · No accounts
Demonstration: the form breathes while idle, reacts while listening, pulses while transcribing and settles when the text is inserted. The audio is simulated on this website.
Good morning
ShellVoice is ready.
Hold⌥ Spaceto talk — anywhere
Recent activity
- “Create a useQuery hook for the organization settings and wire it into the sidebar”
- “Build completed. 142 tests passed.”
- “Run the database migration then restart the dev server”
- “Claude finished the task and needs your attention”
Interactive demonstration. Click the sidebar, click the field, or focus the window and holdSpaceto dictate. Speech and transcription are simulated — nothing is recorded.
The idea
A faster way to interact with your computer.
ShellVoice makes your voice a first-class way in and out of your Mac. Speak, and the words land in whatever app you're using. Ask your tools to speak, and they do.
Speaking…
Transcript inserted where you were typing
Builds, tests and agents speak back
Voice → Text
Dictate into any app. Cleanly.
Speak the way you'd talk to a colleague. ShellVoice turns it into clean text and places it in the app you were working in, with the spacing between sentences handled for you.
Hold ⌥ Space · speak · release
- One shortcut, everywhere
- Hold ⌥ Space in your editor, terminal or browser. Release, and the transcript is inserted where your cursor was.
- Local by default
- Transcription runs on your Mac with whisper.cpp, Metal-accelerated. Audio never leaves the machine.
- Knows your vocabulary
- A custom dictionary fixes technical terms — “use query” becomes useQuery — and biases recognition toward them.
- Never loses a word
- If text can't be inserted, the transcript stays on your clipboard. Toggle mode is there if you'd rather not hold a key.

Global voice
Press a shortcut. Speak. Keep working.
ShellVoice lives in your menu bar. When you hold the shortcut, a small overlay appears above whatever you're doing — fullscreen apps and every Space included — and never takes focus away from your work.

1Hold the shortcut and speak

2Release. Transcription runs locally

3Text lands where you were typing
Default shortcut ⌥ Space · hold to talk, or switch to toggle in Settings
Text → Speech
Natural voices that run on your Mac.
Have a long document read to you, check how a message sounds, or let a script speak up when something finishes. No cloud voices, no accounts, no per-character pricing.
- Free neural voices
- Download Nova, Miles, Lessac, Ryan, Amy, Alba in-app. They're synthesized entirely on-device with Piper.
- System voices too
- Every macOS voice you already have is available alongside them.
- Full playback control
- Play, pause, resume, stop and replay. Speed from 0.5× to 2×, volume to taste.
- Type, then ⌘ ↩
- Paste anything into the Text to Speech screen and hear it read back.

Hear your computer
Don't stare at the terminal. Hear it.
Long builds, test suites, deploys and AI agents all end the same way: with you checking a window. ShellVoice lets them tell you out loud instead.
# from any script, CI hook or agent wrapperpnpm test && shellvoice notify "Build completed. 142 tests passed."shellvoice speak "Claude finished the task and needs your attention"echo "Deploy finished" | shellvoice speakshellvoice CLI is built alongside the app and installed from Settings → Developer.Built-in agent hooks
PlannedAutomatic voice updates from Claude Code, Codex and Gemini CLI sessions are on the roadmap. Today you wire them yourself with one line: whenever your tool can run a command, it can speak.
Developer workflows
Made for people who live in an editor and a terminal.
ShellVoice was built by developers for the way developers work: technical vocabulary, keyboard-first, and nothing that phones home.
A real CLI
speak, notify, status, stop — over a local, user-only socket. Pipe text in from anything.Custom dictionary
Local-first pipeline
Privacy
Your voice stays on your Mac.
Voice apps hear everything you say. ShellVoice is built so that nobody else does.
Local transcription by default
Audio is transcribed on-device and discarded immediately after. It is never written to disk or logged.
No accounts, no analytics, no servers
There's nothing to sign up for and nothing reporting back. The app makes network requests only to download models and voices you choose.
Cloud only if you ask for it
An optional OpenAI-compatible provider sends audio to the endpoint you configure — and only while it's selected.
History you control
Dictations and spoken text are kept in a local file readable only by your user. It's capped, optional and clearable.
Give your Mac a voice.
Dictate anywhere. Hear your tools. Keep everything on your machine.
- macOS
- Available for macOS
- Windows
- Windows planned
Version 0.1.0 for macOS 12.0+ · Signed builds are being prepared