Windows builds ship as a single installer, InkSpoke-win-Setup.exe. There is no MSI and no Store package — the installer also wires up automatic updates.
- 01Download the installerGrab
InkSpoke-win-Setup.exefrom the download page. If the download is slow or blocked on your network, the page also offers a mirror link. - 02Clear the SmartScreen promptCurrent builds are not code-signed, so Windows will show "Windows protected your PC". Choose More info → Run anyway. If that makes you uncomfortable, that is a reasonable instinct — verify the download came from the official release page first.
- 03Let it install and launchInkSpoke installs per-user, so no administrator prompt is needed. It starts in the system tray rather than opening a main window.
ⓘ
The installer is large — roughly 690 MB — because the speech runtime and GPU libraries are bundled rather than downloaded later.
Requirements
- Windows 10 version 1903 or newer, 64-bit
- 4 GB RAM minimum, 8 GB recommended
- An NVIDIA GPU is optional — it enables CUDA acceleration for local transcription, which is dramatically faster than CPU
Permissions
Windows needs far less setup than macOS. Microphone access is governed by the system privacy setting; if it is blocked, InkSpoke will send you to Settings → Privacy → Microphone. Nothing extra is required for typing into other applications.
!
One real limitation: Windows will not let a normal application type into a window running as administrator. If injection silently fails in an elevated app such as Task Manager or an admin terminal, use Restart as administrator from the tray menu.
Where your data lives
windows%APPDATA%\InkSpoke\
That folder holds settings.json, the local inkspoke.db database, and logs. Downloaded speech models and saved recordings default to the same place but can be moved to another drive from Settings → Storage — useful, since the larger models run to several gigabytes.