You think faster than you type. Research from Stanford found that speaking is about three times faster than typing on a phone, and it's easier on your hands and wrists too. That's why voice typing apps like Wispr Flow have taken off. The catch: most of them cost a monthly fee, and every word you say travels to their servers.
Handy is the free, open source alternative. Hold a key, say what you want to write, let go, and the text appears wherever your cursor is: an email, a Slack message, a document, or a prompt for ChatGPT or Claude. It runs on your own computer, so nothing is uploaded and it keeps working with no Wi-Fi. This guide covers what your computer needs, how to set Handy up in a few minutes, the settings worth turning on, and how it compares with the alternatives.
Handy is a free, open source dictation app for Mac, Windows and Linux, made by developer CJ Pais. Its motto is "one tool, one job": turn what you say into text and put it in a text box. It's free under the MIT licence, with no account and no subscription.
It's written in Rust, a fast and modern programming language, which keeps it small and quick. The speech recognition runs through transcribe.cpp, an engine made by the Handy team that runs dozens of speech models, including OpenAI's Whisper and NVIDIA's Parakeet, on your own computer. It's good enough that other dictation apps use it too, including VoiceInk on the Mac.
Because the speech model runs on your own computer, your computer does the work. The good news: speech models are far smaller than chat AI models, so you don't need a powerful machine. Handy doesn't publish official minimum specs, so here's our guide based on the size of each model:
Handy scores its own Whisper models for accuracy and speed. Here they are out of 10, with the download size:
One tip: Handy can unload the model from memory when you haven't used it for a while, so it isn't holding on to memory all day.
Prefer the terminal? Handy is also available through these package managers. They're kept up by the community rather than the Handy developers:
# Mac (Homebrew)
brew install --cask handy
# Windows (winget)
winget install cjpais.Handy
# Debian or Ubuntu, after downloading the .deb
sudo apt install ./Handy_*.debStart with the one Handy marks as recommended, a Parakeet model that runs well on an ordinary laptop. It's English only, so if you speak another language, pick a model that supports it: the Whisper models cover the most languages. Handy downloads the model once, which takes a minute or two, and after that everything works offline.
Choose a keyboard shortcut in the settings. Then click into any text box, hold the shortcut, say what you want to write and let go. Your words appear a moment later. Prefer not to hold a key? Tap it once to start and again to stop.
Handy looks simple, but there's more inside. Most of these live under Advanced in the settings.
All four turn your voice into text. They differ in privacy, polish and cost.
Best forFree, private voice typing on any computer
Hold a key, talk, and the text appears in any app. Runs on your computer, light and quick. Text comes out as you said it unless you switch on AI cleanup.
Best forThe most polished, if you don't mind paying
Very smooth, and it tidies your words with AI automatically. Not open source, and your audio is processed in the cloud.
Best forA polished Mac app with modes for each app
A native Mac app that runs offline, adjusts to the app you're in and has a voice assistant. Free if you build it yourself, otherwise a paid licence. Needs macOS 15 or later.
Best forEverything in one: dictation, meetings, notes
Dictation plus meeting transcription with speaker names, a notes app and an AI agent. Local or cloud models. Heavier, and some features need an account or the paid plan.
A quick way to choose: want polish and don't mind paying or the cloud? Wispr Flow. On a Mac and want the slickest open source app? VoiceInk. Need meeting notes too? OpenWhispr. Want free, private voice typing on any computer? Use Handy.
You may also see FreeFlow recommended. It's a free, open source Mac app with excellent AI cleanup, but by default it sends your voice to Groq's servers, so it isn't private unless you point it at a model on your own computer.
How we score: each open source app gets up to 2 points in five areas, for a total out of 10. Privacy: does your voice stay on your computer by default? Cost: is it free, with nothing important behind a paid plan? Platforms: does it run on Mac, Windows and Linux? Ease: how quickly you're set up, and how good the text is straight away. Features: what it can do beyond plain typing.
| App | Score | Privacy | Cost | Platforms | Ease | Features |
|---|---|---|---|---|---|---|
| Handy | 8.5/10 | 2 | 2 | 2 | 1 | 1.5 |
| VoiceInk | 8/10 | 2 | 1 | 1 | 2 | 2 |
| OpenWhispr | 7.5/10 | 1.5 | 1.5 | 2 | 0.5 | 2 |
| FreeFlow | 6.5/10 | 0.5 | 1.5 | 1 | 2 | 1.5 |
A note on these scores: we only score open source apps, which is why Wispr Flow isn't included. The scores are our opinion, based on each app's features, platforms, licence and what users report, not side-by-side lab tests. Apps change quickly, so check each one's latest version before you decide.
Give it a day. Write your emails and messages by talking, and most people find they don't want to go back. If you'd like the AI cleanup to stay on your computer too, our guide to running AI with Ollama shows you how in about five minutes.
And if you build things: Handy's speech engine, transcribe.cpp, is open source too, under the MIT licence. It can power your own tools, like a meeting notetaker that never uploads a call. That's a guide for another day.
For most people, Handy. It's free, open source, works on Mac, Windows and Linux, and runs the speech recognition on your own computer. On a Mac, VoiceInk is also excellent but its ready-made app is paid. If you need meeting notes too, look at OpenWhispr.
Yes. Handy is open source under the MIT licence. There's no account, no subscription and no word limit. The only optional cost is if you turn on AI cleanup with a paid online provider. With a local model through Ollama, that's free too.
No. The speech recognition runs on your own computer, so your recordings stay with you and it works without internet. It only goes online to download a speech model the first time, check for updates, and send text to an AI provider if you choose to turn on AI cleanup.
Handy doesn't publish official minimum specs. As a guide, any computer from the last few years with 8 GB of RAM runs the recommended Parakeet model or Whisper Small comfortably, with no graphics card needed. For the larger Whisper Turbo and Large models, 16 GB of RAM and a graphics chip (any Apple Silicon Mac, or a graphics card on Windows or Linux) make them much faster. Each model takes 0.5 to 1.6 GB of disk space.
For turning speech into text, it's very close, and it's free and private. Wispr Flow is more polished at tidying your words with AI automatically. Handy can do that too, but you have to switch it on and connect an AI provider yourself.
Start with the one Handy marks as recommended, a Parakeet model that runs well on an ordinary laptop processor. If you speak a language other than English, check that the model supports it: the Whisper models cover the most languages. If you have a good graphics chip and want the best accuracy, try Whisper Large.
Yes. Handy runs on Windows, macOS and Linux. On Linux, install xdotool (X11) or wtype or dotool (Wayland) so it can type into other apps.
No. Handy is a desktop app. On a phone, use the dictation built into your keyboard.
Free, private voice typing for your computer. Press a shortcut, speak, and your words appear in any app. Works offline on Mac, Windows and Linux.
Run AI chat models like Gemma, Qwen and DeepSeek on your own computer with one command, for free and in private. Other apps can use them too.
