Handy

1. What is Handy?

Handy is a free desktop app (Windows, macOS, Linux) that lets you talk instead of type. You press a shortcut, speak, and let go. A moment later your words appear as text in whatever box you were typing in: a document, a browser, a chat, an AI prompt. The speech recognition runs on your own computer, so according to the project your voice is not sent to the cloud.

It stands in for paid dictation services. Handy is open source (MIT licence) and has no paid tier inside the app.

A good fit if:

• You type slowly or your hands get tired and you would rather dictate.
• You do not want to upload your voice to an online service.
• You want something simple: one shortcut, one job.

Not a fit if:

• You only have a phone. Handy is a desktop app with no Android or iPhone version.
• You need meeting transcripts with speaker labels or subtitle files for video. Handy only types what you say into a text box.
• You need perfect accuracy for every accent. Quality depends on the model you pick, so test it yourself (see section 4).

2. What you need

• A computer running Windows, macOS or Linux. The project ships installers for Windows (x64 and arm64), macOS (Intel and Apple chips) and Linux (.deb, .rpm, AppImage).
• A microphone. The one built into a laptop is enough. With Bluetooth headphones on a Mac, see the troubleshooting section.
• Disk space for the speech model. On first use you download a "model", the data file that lets the computer recognise speech. The project lists these Whisper model sizes: Small about 487 MB, Medium about 492 MB, Large about 1.1 GB, Turbo about 1.6 GB. It also offers a Parakeet model optimised for CPUs, which works well even without a graphics card. Keep a few GB free.
• Internet is needed only to download the app and the model the first time. After that the project says it works fully offline.
• No account, no API key, no cost.

3. Step-by-step install

Step 1: Download the installer

Go to the official site https://handy.computer or the project's Releases page https://github.com/cjpais/Handy/releases and pick the file for your machine (the latest version at the time of writing is v0.9.7):

Your machine	File to download
Regular Windows PC (Intel/AMD)	the file ending in x64-setup.exe (or x64_en-US.msi)
Windows on ARM	the file ending in arm64-setup.exe
Mac with Apple chip (M1, M2, M3...)	the aarch64.dmg file
Mac with Intel chip	the x64.dmg file
Ubuntu/Debian	the .deb file

Not sure which you have? Most Windows PCs are the regular x64 kind. On a Mac, click the Apple menu and choose "About This Mac": if you see "Apple M...", it is an Apple chip.

Step 2: Install the app

Windows: double-click the downloaded .exe and follow the screens. If Windows warns you because the app came from outside the store, make sure you downloaded it from one of the two official pages above before you allow it.

macOS: open the .dmg and drag Handy into the Applications folder.

Ubuntu/Debian: the project recommends installing the .deb with APT so required libraries come along. Open a Terminal in the folder with the download and run:

sudo apt install ./Handy_*.deb

Avoid dpkg -i unless the dependencies are already installed. If you already used it and hit errors, run sudo apt --fix-broken install.

Quick install by command (optional). The README lists two community-maintained packages, not maintained by the Handy developers: on Mac brew install --cask handy, on Windows winget install cjpais.Handy.

Step 3: Open Handy and grant permissions

Open Handy. The operating system will ask for microphone and accessibility access. Accessibility is what lets Handy paste text into the box you are typing in. Allow both. On macOS, if you refused by mistake, turn it back on in System Settings under Privacy & Security.

Step 4: Choose a speech model

Open Handy's settings (the Models section) and download a model. On a weak machine or one without a graphics card, try the Parakeet model (per the project, optimised for CPU with automatic language detection). With a graphics card, the Whisper models use GPU acceleration where available. When the download finishes, the model shows as "Downloaded".

Step 5: Set your shortcut

In Settings, pick the shortcut you like. There are two styles: hold the key to talk and release to stop, or tap once to start and tap again to stop. Choose a combination that does not clash with other software.

4. First use

1. Open any app with a text box, for example Notepad or Notes.
2. Click in the box so the cursor blinks.
3. Hold your shortcut, say one short, clear sentence, then release.
4. Wait a moment. The text appears in the box.

If nothing appears but you can see the transcript in Handy's History, check the paste problem below.

About other languages: the project describes Whisper as a multilingual family of models and Handy lets you change the language in settings, but the README does not promise accuracy for any particular language. Try a few sentences with a Medium or larger Whisper model and keep the one that suits your voice.

5. Common problems and fixes

Symptom	Cause	Fix
Pressing the shortcut does nothing	Microphone or accessibility permission missing	Open your system's privacy settings, enable microphone and accessibility for Handy, then reopen the app
The transcript is right in History but old copied text gets pasted	The receiving app reads the clipboard after Handy has already restored it (tracked by the project as issue #502)	Press Ctrl+Shift+D (Mac: Cmd+Shift+D) to reveal Debug, try "Reliable Paste (Beta)" on Windows/macOS, or raise "Paste Delay (After)"
Sound gets quiet on a Mac with Bluetooth headphones while recording	Bluetooth switches to two-way audio when the headset mic is used	Keep the headphones as output and pick the Mac's built-in or an external mic in Handy
A shortcut with the fn/Globe key does nothing on a Mac	That key only works from Apple keyboards	Use ctrl, option, shift, command combinations instead
Linux: text is not typed into the box	Missing input tool for your display server	X11: sudo apt install xdotool. Wayland: sudo apt install wtype (or dotool)
Linux: will not start, mentions libgtk-layer-shell.so.0	Missing runtime library	Ubuntu/Debian: sudo apt install libgtk-layer-shell0
Model will not download (company network, proxy)	Network blocks the automatic download	Follow "Manual Model Installation" in the official README: download the model file in a browser and put it in Handy's models folder

6. Uninstall and update

• Update: download the new version from the Releases page and install over the old one. Note that the Homebrew and winget packages are community-maintained and may lag behind.
• Uninstall: remove it like any other app (Windows: Settings > Apps; Mac: drag Handy from Applications to the Trash; Linux: use your package manager).
• Remove leftover data (downloaded models, history): per the README, the data folder is ~/Library/Application Support/com.pais.handy/ (macOS), C:\Users\<your name>\AppData\Roaming\com.pais.handy\ (Windows) or ~/.config/com.pais.handy/ (Linux). Delete it for a clean wipe.

7. FAQ

Does Handy cost money? No. It is free and open source under the MIT licence.

Do I need internet? Only to download the app and the model the first time. Afterwards recognition runs on your machine.

Is my voice sent anywhere? According to the project, processing is entirely local and nothing goes to the cloud.

Will it run on a weak computer? The project offers a CPU-optimised Parakeet model and several Whisper sizes. On a weak machine pick a smaller model and test the speed yourself.

Does it work on a phone? No. Handy is for Windows, macOS and Linux.