<0.5s on your GPU
No upload. No queue. No waiting on someone else’s servers.
v1.3.0 · free · local · open source · Mac + Windows preview
On-device dictation in <0.5s — not a ~$2B cloud bill at ~2–3s. Your Mac. Your voice. $0 forever.
Live race
<0.5s local vs ~2–3s cloud
Versus
Same job. Better defaults. No monthly tax.
Benefits
Less cloud tax. More Mac. One job done well.
No upload. No queue. No waiting on someone else’s servers.
Hold. Speak. Release. Text lands where you are.
Forever free. MIT. No trial cliff.
Built for Apple Silicon. Windows 10/11 x64 gets a CPU preview with the same flow: hold Alt Space, talk, let go.
No account. Audio doesn’t leave your Mac for local ASR. Telemetry is opt-in, anonymous, and never sees your words — read the contract.
Check for Updates… built in (Sparkle). No re-downloading DMGs like it’s 2009.
The math
Setup · v1.3.0
v1.3.0: one on-device model (Parakeet Redux), nothing to configure. Signed and notarized by Apple — it opens like any other app.
MacWispr 1.3.0 (DMG) → drag MacWispr into Applications.
Open MacWispr from Applications. The first launch sets up the speech model (a one-time download, about a minute on a good connection).
Allow Microphone and Accessibility, then hold Option Space, speak, and let go.
Preview · Windows 0.1.0
Preview: Windows 10/11 on x64. Speech runs on your CPU with Parakeet v3, so no GPU is needed. The installer isn’t code-signed yet, so SmartScreen may ask first.
MacWispr for Windows 0.1.0 (installer) → if SmartScreen appears, choose “More info”, then “Run anyway”. No admin rights needed.
MacWispr sits in the system tray. The first launch downloads the speech model (about 670 MB, one time).
Allow microphone access if Windows asks, then hold Alt Space, speak, and let go.
Open-source dictation for Mac, now in preview on Windows. No account. No $2B subscription story.