Open-source speech-to-text on Windows with Cursay

Cursay has a native Windows app for system-wide push-to-talk dictation. The app uses a local Whisper service and keeps history on your device. Windows support currently targets x64 Windows 10 version 1809 or newer and Windows 11.

Published and reviewed · Cursay product guide

Check build availability before installing

The current GitHub v1.2.1 release has an Ubuntu archive and no Windows installer asset. Windows app source and installer build scripts are available. If a later release includes Cursay-<version>-windows-x64-setup.exe, follow that release’s instructions; this guide currently documents building from source.

Install the development requirements listed in the Windows source guide: .NET 8 SDK, Python 3.12 x64 and Inno Setup 6. From PowerShell, clone the repository and run the build script. It runs the app tests, packages the local speech backend, creates a per-user installer and writes a checksum under windows\artifacts.

git clone https://github.com/shadoprizm/cursay.git
cd cursay
.\windows\scripts\build.ps1

Set up microphone access and local recognition

Run your built installer and open Cursay from the notification area. If recording is blocked, check Settings → Privacy & security → Microphone in Windows 11, including Microphone access and Let desktop apps access your microphone. Windows 10 uses the Privacy → Microphone settings page.

Select the local transcription option. The first dictation downloads the English base.en Whisper model, roughly 150 MB. After that download, local speech recognition can run offline. Settings, models, history and recordings you explicitly preserve are stored under %LOCALAPPDATA%\Cursay.

Use push-to-talk in a text field

Ctrl + Space must be available as a global shortcut. Another program can reserve it. Check microphone input, the selected endpoint and the clipboard result when troubleshooting rather than assuming the shortcut is the only cause.

  1. Focus the text field where you want the transcript.
  2. Hold Ctrl + Space and speak.
  3. Release the shortcut to finish recording. Check the result at the cursor, or paste it manually from the clipboard.

Understand the current Windows limitations

The Windows build is x64-only. Release signing is not yet configured in the Windows build guide. This guide does not instruct you to bypass Windows security warnings.

Cursay cannot automatically paste into an elevated application when Cursay runs without elevation; the transcript remains on the clipboard. Test in an ordinary text editor first.

Windows currently has a local/custom transcription path but lacks the account-linking and token-refresh integration present on Mac and Ubuntu. Do not buy Pro expecting the same linked cloud workflow on Windows today. Custom remote endpoints have their own privacy and billing terms.

Common questions

Is local Windows dictation free?

Yes. Cursay’s local Whisper path is open source, unlimited and accountless after installation.

Where can I get a Windows installer?

Check the assets on the GitHub releases page. The current v1.2.1 release has no Windows installer; use the Windows source build guide if you are comfortable building it.

Sources and next steps

This guide follows the Cursay source setup guide and published release assets. Availability may change; check those sources before installing.

Review privacy, cloud subprocessors and pricing before choosing a remote provider.

Open Cursay on GitHub