WinSTTInstall
Download the right WinSTT package for macOS, Linux, or Windows. No Python, no account, no separate speech server.
WinSTT is a self-contained desktop app. Download it, open it, and start talking. There is no Python to set up, no separate speech server to run, and no account needed for transcription on your own machine.
Two ways to get it on Windows
Pick whichever feels easier - they are the same app, just delivered differently. The portable version is a folder you unzip and run, with nothing added to your system. The installer sets WinSTT up like any normal program, complete with a Start-menu shortcut. You can switch between them anytime.
- Adds a Start-menu shortcut
- Opens automatically when it's done
Not sure which to choose?
Go with the installer if you just want WinSTT on your PC the usual way. Pick the portable version if you'd rather not install anything - for example on a work computer, or to carry it on a USB stick.
Current alpha downloads
Use the download menu to get the current package for your desktop OS. WinSTT detects macOS, Linux, or Windows and only lists the matching binaries.
Which package?
| Platform | Artifact | Use when |
|---|---|---|
| macOS Apple Silicon | WinSTT_<version>_aarch64.dmg | You have an M-series Mac. Intel macOS builds are not published in this alpha because the current native dependency stack does not provide the needed x86_64-apple-darwin prebuilts. |
| Linux x64 | .AppImage / .deb / .rpm | Use AppImage for a portable package, deb for Debian/Ubuntu-style systems, or rpm for Fedora/RHEL-style systems. |
| Windows x64 | WinSTT.exe / WinSTT-portable.zip | Use the executable for the normal Windows alpha, or the zip when you want to keep the app folder self-contained. |
System requirements
| Requirement | Minimum | Notes |
|---|---|---|
| macOS | Apple Silicon Mac | Current alpha publishes arm64 DMG only. |
| Linux | x86_64 desktop distro | Use the AppImage, deb, or rpm package that best matches your system. |
| Windows | Windows 10 1903+ or Windows 11, x64 | The Windows build can use DirectML on D3D12-capable GPUs and falls back to CPU. |
| CPU/RAM | Modern 64-bit CPU, 4 GB RAM | Tiny/base models are light; larger Whisper or NeMo models benefit from 8 GB+. |
| Disk | ~1 GB + model cache | The app includes a starter model; additional model downloads are cached locally. |
| Microphone | Any input device PortAudio can open | Listen mode needs a loopback/monitor device for system audio. |
Run it
Download the package
Use the OS-aware download menu above and choose the package type that fits how you want to run WinSTT.
Launch the app
On macOS, open the DMG and launch WinSTT. On Linux, run the AppImage or install the deb/rpm with your package manager. On Windows, run
WinSTT.exeor unpackWinSTT-portable.zipand launch the executable inside.Handle first-run OS prompts
Alpha packages may trigger platform trust prompts until signing and notarization are fully wired for every OS (see the note below).
Onboard and pick a model
The first-run wizard asks for a microphone, a hotkey, and a local-vs-cloud preference. The starter model is bundled, and additional models download from the UI without restarting the app.
Confirm the source before you continue
Alpha packages may trigger platform trust prompts until signing and notarization are fully wired for every OS. Only continue past an OS warning after you have confirmed the asset came from the WinSTT GitHub Release.
Verification
The current alpha release publishes downloadable packages but not updater metadata or detached signature sidecars. For a byte-level check, compare the file's SHA-256 digest with the digest shown in the GitHub release asset metadata, or use the workflow run attached to the release as the build provenance.
Verify downloads
Compare release asset digests and understand the current alpha signing status.
Quick start
Finish onboarding and dictate your first sentence.
Choose a model
Browse 70+ STT models and pick the right accuracy/speed trade-off.
Troubleshooting
Blank window, slow transcription, audio devices, and model-download fixes.
Quick Start
Download, launch, and dictate your first sentence in under two minutes — no Python, no accounts, no setup.
Dictation
Press your hotkey (or just speak) and WinSTT drops polished text at your cursor. The four modes — Push-to-Talk, Toggle, Listen, and Wake Word — decide what starts a recording; everything after is identical.