Macros, not single shortcuts
Hold a key down for a set duration, then release it. Chain any number of steps with delays you control. Mix keys, combinations, clicks, movement, scrolling and typed text in one command.
▶ IN DEVELOPMENT // macOS + WINDOWS + LINUX // COMING TO STEAM
VoiceCmd
Talk. It obeys.
Voice macros with real key holds.
Say “reload” and the R key happens. Say “unleash the kraken” and a whole macro happens — a sequence of keystrokes with real holds and precise delays, mouse clicks, typed text, exactly as you defined it. Recognition runs entirely on your own machine: no account, no subscription, and no audio ever leaves your computer.
VoiceCmd is going out on Steam. The app is registered, its Windows, macOS and Linux depots are set up and the store art is drawn — what is missing is a published store page, which is why there is no button here yet. There will be one the day that page is public, and not a day before it.
▶ SPEC SHEET
WHAT IT DOES
Most voice tools stop at a single shortcut. VoiceCmd is built for everything past that point.
The whole keyboard
F1 through F24, the numpad, the navigation cluster, media keys and chorded combinations — delivered as real input, not as text pasted into a box.
Listens when you want it to
Push-to-talk or a toggle hotkey, so it will not fire mid-sentence while you are on a call. An optional wake word if you would rather address it by name. The hotkey gates matching rather than the microphone, so there is no spin-up delay.
Profiles that follow you
Group commands per application. VoiceCmd notices which app you switch to and loads the matching profile automatically. Export any profile as JSON to back it up or share it.
Fast, and honest about it
Roughly half a second from the moment you stop speaking to the keystroke landing. A built-in debug view breaks the timing down step by step, so you can check that claim rather than take our word for it.
Understands you, not just your diction
Fuzzy matching tolerates typos and homophones. Longest phrase wins, so “reload weapon” never fires “reload” by mistake. Per-command and global cooldowns stop one utterance firing twice.
▶ ON SCREEN
WHAT IT LOOKS LIKE
▶ YOUR DATA
IT NEVER LEAVES YOUR MACHINE
The speech model is bundled inside the app and runs on your own computer. Audio is processed as a live stream, never written to a file, and discarded the moment it becomes text. There are no accounts, no telemetry and no servers — VoiceCmd works with the network switched off. Read the full privacy policy.
▶ SUPPORT
SOMETHING NOT WORKING?
Email info@machinautstudios.com and we will help. A real person reads it.
Tell us this and it will be quicker
Open Settings → Debug Info and copy the block — it holds the version, the exact build, and your operating system. Select it, copy it, and paste it into the email.
Nothing is being typed (macOS)
macOS silently discards simulated keystrokes until the app is allowed to send them. Grant VoiceCmd Accessibility in System Settings → Privacy & Security. The app shows an INPUT BLOCKED chip while that permission is missing.
It is not hearing you
Open Settings → Microphone input and watch the level meter while you speak. If it never moves, the app is being handed silence — pick the correct input device there, and check the microphone is not muted in your system sound settings.