VocaWin
download the beta
private speech-to-text, for a Windows PC you own
voice typing that stays on this PC.
file an issue source nightly from main
Beta. Unsigned. Windows will likely say the publisher is unknown. More info, then Run anyway. Latest tagged build is on Releases.
Windows 10 / 11
this PC
The point is ordinary Windows work. Hold a hotkey, talk, and the transcript is meant to land where the caret already is.
the intended gesture
VocaWin is a tray app for Windows 10 and 11. Hold the hotkey (default: Right Alt), speak, and the app injects text through SendInput. Browser, editor, chat, terminal: wherever Windows will accept keystrokes.
on this PC
After you download a Whisper model, recording and speech-to-text stay on this Windows machine. That is on-device processing, not a Voca cloud. A network is needed for the first model download, then dictation can work offline.
the family
VocaLinux, VocaMac, and VocaPhone already exist for other machines. VocaGateway is optional self-hosted compute for those clients. VocaWin does not offer a gateway mode today. The shared home is vocahq.com.
This is the flow in the unsigned beta. Download a model once, hold the hotkey, and text is meant to land at the caret. It is still tester-only, not a store listing.
A Whisper model is meant to live in a local VocaWin folder on this PC. After that file is on disk, transcription does not need the internet.
Hold Right Alt to record, or use the double-tap toggle. The tray icon shows idle, recording, and processing.
The transcript is meant to appear in the focused field. Elevated admin windows may refuse SendInput. That limitation is real in this beta.
honest status: GitHub Releases has an NSIS current-user
setup.exe and an MSI. Both are unsigned. Windows will
likely say the publisher is unknown. More info, then Run anyway.
There is no Microsoft Store listing and no auto-update.
Read the setup guide
first, and
file an issue
if something breaks.
The beta records with WASAPI on this machine and runs a local speech-to-text model. That is on-device. It is not self-hosted gateway mode, and it is not a hosted speech API.
Audio and the speech-to-text model stay on the Windows computer after the model download. Dictation can then work offline.
Other Voca apps can send audio to a gateway you run yourself. VocaWin does not expose that path today. A gateway is still not a Voca cloud.
You can install a tester build from GitHub Releases. Treat it as early. It is not signed, not on the Microsoft Store, and not a stable public release.
today
known limits
Start at vocahq.com if you want the whole map. Each product keeps its own install path and honest status.
family home
vocahq.com Private speech-to-text across the machines you own.
Linux
available now
System-wide voice typing for real Linux desktops. On-device speech-to-text on X11 and Wayland.
macOS
beta
A native menu bar app for Apple Silicon. Hold a hotkey, speak, and text appears at the cursor through an on-device speech-to-text model.
iPhone + Android
beta / testflight
A private voice keyboard for phones. Android 13+ has a public beta. iPhone is a public TestFlight beta for iOS 17+, capped at 1,000 seats, or build from source.
infrastructure
early
Self-hosted shared compute for Voca clients. Pair with a QR code. Use a trusted LAN, a private encrypted network, or HTTPS.
VocaWin is a native Windows app built in the open: tray icon, hotkey, local Whisper model, text at the caret. No Voca account. No hosted speech service. The unsigned beta lives on GitHub Releases. This is still a tester build, not a signed store ship.
download the beta from Releases ↗Yes. An unsigned beta is on GitHub Releases. Windows will likely warn that the publisher is unknown. That is SmartScreen doing its job. More info, then Run anyway if you trust the file. This is not a Microsoft Store listing and not a signed public release. Read the setup guide first, and file an issue if something breaks.
Yes. A moving unsigned build from
main is on the
nightly
GitHub Release. Same SmartScreen story as the tagged builds.
Prefer the latest tagged Release if you want the last cut we
named. Nightly is for testers who want today's tree.
That is the design. After a Whisper model is downloaded, recording and speech-to-text run on this Windows machine. No Voca cloud, no Voca account, no hosted speech API. The first model download does need a network. Treat the beta as early.
Not for VocaWin. VocaWin is an on-device Windows app. VocaGateway is optional self-hosted infrastructure for other Voca clients. Gateway mode is not on-device: configured audio travels to the host you chose.
VocaHQ is the family home for VocaLinux, VocaMac, VocaWin, VocaPhone, and VocaGateway. The directory, privacy paths, and project status live on vocahq.com.
The beta defaults to Right Alt for push-to-talk, with an optional double-tap toggle. You can change the hotkey in Settings. That default can still move.
Support follows the Whisper model you download. The first-run
default is the small English tiny.en model, with
language set to automatic. We list only what the selected model
can actually transcribe.
Yes. The repository is licensed under AGPL-3.0-or-later, stays free, and does not require a Voca account. Sibling Voca apps also use AGPL-3.0.
unsigned beta
Grab the beta from GitHub Releases if you want to poke at this PC. If you need a quieter desk, start at vocahq.com and pick a machine that already ships.
