GatedVoice is a free and open-source alternative to paid dictation apps like Wispr Flow. Speech-to-text runs 100% on your own machine with Whisper — no account, no subscription, no per-word cost, and nothing ever leaves your computer.
A ShyLabs project.
Good dictation apps cost money and send your voice to their servers. GatedVoice does neither.
- 🆓 Actually free — no trial, no subscription, no usage cap. MIT licensed.
- 🔒 Private by design — audio is transcribed locally with Whisper and never uploaded. It works with your Wi-Fi off.
- ⌨️ Works everywhere — hold a hotkey in any app (browser, email, Slack, code editor, Word) and your speech is typed where your cursor is.
- ⚡ Fast — the default model loads in about a third of a second and transcribes in near real time.
- 🧠 Smart — a custom dictionary that fixes how your names and jargon are spelled, text snippets, usage insights, and optional AI cleanup.
- 🪶 Tiny footprint — a single native app that lives in your system tray.
- Grab the latest Release and download the ZIP for your PC:
GatedVoice-win-x64.zip— most Windows PCs (Intel / AMD)GatedVoice-win-arm64.zip— Windows on ARM (Snapdragon, etc.)
- Unzip it anywhere and run
GatedVoice.exe. - On first launch GatedVoice downloads the speech model once (~60 MB), then prepares a smaller live-preview model (~32 MB) in the background. A "GatedVoice is ready" popup appears as soon as normal dictation is available.
- Click into any text box, hold
Ctrl+Space, talk, and release. Your words are typed in.
Windows may show a SmartScreen warning because the app isn't code-signed (signing costs money; this is free). Click More info → Run anyway.
You need the .NET 8 SDK.
git clone https://github.com/yzgershon/GatedVoice.git
cd GatedVoice
# run it
dotnet run -c Release
# or build a standalone app for your PC (win-x64 or win-arm64)
dotnet publish -c Release -r win-x64 --self-contained true -o dist| Hotkey | What it does |
|---|---|
Hold Ctrl + Space |
Dictate. Speak while held, release to insert the text. |
Hold Alt + Space |
Dictate and clean it up with AI (see Transforms). |
Both are push-to-talk — text lands the instant you let go. Everything else lives in the tray icon (bottom-right of your taskbar): enable/disable, insights, launch-at-login, and quick access to your settings.
- Custom dictionary — teach GatedVoice to spell names and jargon right. It hints Whisper toward your spelling and fixes common mishears.
- Snippets — say a trigger phrase, get a saved block of text (your email, a signature, a link).
- Insights — total words, average WPM, words per day, and a daily streak.
- Style controls — trailing spaces, filler-word removal ("um", "uh"), paste vs. type.
- AI Transforms (optional) —
Alt + Spaceruns your dictation through an LLM to fix grammar and punctuation while keeping your voice. Use Ollama locally for free, or plug in an Anthropic/OpenAI key. Off by default. - Rebindable hotkeys, swappable Whisper models (base → small → medium for more accuracy), and a built-in scratchpad.
Your voice is turned into text on your device by Whisper. No audio and no transcripts are ever sent anywhere — the only network requests GatedVoice makes on its own are the one-time speech-model downloads (and optional cloud AI, only if you turn it on and add a key). Your settings and history live in a local folder (%APPDATA%\GatedVoice) that only you can see.
| Piece | Tech |
|---|---|
| Speech-to-text | Whisper.net (whisper.cpp), native, offline |
| Audio capture | NAudio (WASAPI microphone) |
| System-wide hotkey + text insertion | Low-level keyboard hook + clipboard paste |
| App + UI | C# / WinForms + WebView2 dashboard (.NET 8) |
Built for Windows on x64 and ARM64, running natively on both.
Issues and pull requests are welcome. GatedVoice started as a personal project to avoid paying for dictation — if it's useful to you, a ⭐ helps other people find it.
MIT — free to use, modify, and share.
Whisper models are © their respective authors and downloaded from Hugging Face on first run.