BUILT AT VEY LABS
LIVE

Neo — voice control that stays on your machine.

A voice-controlled AI interface for macOS, for people whose hands shouldn't be on a keyboard. Speech, intent, and action all resolve locally — sub-100ms on critical commands, nothing leaving the device by default.

neo · listening · offline
84ms
Median wake-word to executed command.
186ms
p95, including multi-step commands.
0 bytes
Audio leaving the device on the local tier.
11 hrs
Idle listening on battery, measured not estimated.
Measured on M2 MacBook Air, wake word to executed action, no network.
WHY IT EXISTS

It started with an injury.

Neo began because typing became painful. RSI takes the keyboard away gradually, and every existing option asked for a trade: dictation that guessed wrong, or cloud assistants that shipped a live microphone feed off the machine in exchange for being useful.

So we built the version we needed: fast enough to replace a keyboard shortcut, private enough to leave on all day, and accurate on the vocabulary of actual work. People with RSI were first. People who think faster than they type came next.

macOSOn-device AIRSI-friendlyNo account required
WHAT IT DOES

Built for a working day,
not a demo video.

Commands, not dictation.

Say what you want done. Neo resolves intent to an action in the app you are in, rather than typing words and hoping.

Local by default.

Speech recognition and intent parsing run on your machine. Cloud models are opt-in, per workspace, and visibly labelled.

Faster than the shortcut.

If voice is slower than the keyboard, nobody uses it. Sub-100ms on critical commands is the design constraint.

Built for whole days.

Warm audio pipeline, low idle draw, no fan. It sits in the menu bar and stays out of the way.

Correctable, not chatty.

Wrong result? One word fixes it. Neo learns your vocabulary — project names, people, jargon — locally.

Works when the wifi does not.

Planes, trains, hospital basements. The local tier has no network dependency at all.

SPECIFICATIONS
PLATFORM
macOS 14+ · Apple silicon
MODELS
On-device speech and intent models, quantized. Optional cloud models per workspace.
PRIVACY
Local tier processes and stores everything on device. No telemetry on audio content.
ACCESSIBILITY
Full keyboard alternative for navigation, editing, and app control. Designed with RSI users.
PRICE
Free local tier. Paid tier for cloud models and team workspaces.
STATUS
Live and shipping · tryneo.dev

Try Neo free.

Voice-controlled AI on your machine. No cloud required, no account for the local tier.