1. Wispr’s servers or your own processor: where each app hears you
This is the difference that decides most comparisons. Wispr’s Data Controls page, last updated August 18, 2026, says transcription always occurs in the cloud, which Wispr calls the best way to provide accurate, low-latency results, and its help center says an internet connection is required for transcription. Flow also uses the name of the app you are dictating into, and the text already in the box, to pick tone, capitalization, and punctuation.
Wispr’s privacy settings are three separate switches. “Improve the model for everyone” lets Wispr use your audio, transcripts, and edits to train its models; the help center says it is on by default for Free and Pro, appears during onboarding, and can be turned off, while Growth and Enterprise keep it off. Dictation Cloud Storage decides whether transcripts, audio, and history are kept on Wispr’s servers. Whatever you choose, Wispr says it may collect usage statistics such as word counts, encrypts data in transit and at rest, and holds zero-data-retention agreements with its third-party AI providers.
Handy’s README describes the opposite design: press a shortcut, speak, and the words appear in any text field, transcribed on your own computer without sending information to the cloud. Silero voice activity detection filters out silence, and you pick the model: Whisper Small, Medium, Turbo, or Large, with GPU acceleration when available, or Parakeet V3, which the README calls CPU-optimized with automatic language detection. Once a model file is on disk, the project describes Handy as working completely offline.
- Wispr Flow: no offline transcription; an account on every device.
- Handy: the network is needed only to fetch a model, and the README documents copying model files in by hand.
- Handy’s source is public, so its behavior can be audited rather than taken on trust.
2. What each costs as of October 2026, and what the licence allows
Wispr’s pricing page, checked October 3, 2026, lists Free at $0 with a cap of 2,000 words a week on desktop and 1,000 on mobile, and Pro at $15 per user per month, or $12 billed annually, with unlimited dictation. Growth and Enterprise plans add team administration, single sign-on, and compliance controls. Students and educators with an education email get 50% off Pro, and Wispr says you can cancel Pro at any time. There is no one-time or lifetime plan, so a year of Pro costs $180 paid monthly or $144 billed annually.
Handy costs nothing and has no paid tier. Its code is under the MIT License, so you may read, change, and redistribute it; the README adds that the Handy name, logo, and brand assets are not open source, so forks must use their own branding. Development is supported by sponsors and donations. The trade-off for free is support: problems go to GitHub issues or the project’s Discord, and the README keeps a candid list of known issues, from Bluetooth headset quality on macOS to Wayland limits on Linux.
- Wispr Flow Free: 2,000 desktop words a week; Pro removes the cap.
- Wispr Flow Pro for a year: $180 monthly or $144 annually.
- Handy: $0, no word limits, MIT-licensed code.
3. Phones, Intel Macs, ARM laptops, and Linux packages
Wispr Flow runs on Mac (macOS 12 Monterey or later, Apple silicon or Intel, about 500 MB free), Windows 10 or 11 on an x64 processor (about 200 MB free), iPhone on iOS 18.3 or later, and Android 13 or newer. Its system-requirements article says iPad, Linux, Chromebooks, virtual machines, and remote desktop environments are not supported, and Windows on ARM and Snapdragon PCs get an “Unsupported Windows Architecture” message. You sign in with Google, Apple, Microsoft, single sign-on, or email.
Handy has no phone app, but its desktop coverage is wider. Version 0.9.8 on the download page offers Mac builds for Apple silicon and Intel, Windows installers for x64 and ARM, and Linux AppImage, .deb, and .rpm packages for both x64 and ARM. It is also published as a Homebrew cask and on winget, which the developers note they do not maintain. On Debian and Ubuntu the README recommends sudo apt install ./Handy_*.deb so that APT installs the dependencies too.
4. Hotkeys, cleanup, and teaching each app your vocabulary
Flow’s default hotkey is Fn on a Mac, or Ctrl+Option on keyboards without Apple’s Fn key, and Ctrl+Win on Windows; on iPhone you tap the Flow keyboard’s microphone and on Android you use an on-screen bubble. Flow cleans up what you say before inserting it and offers a hands-free mode, a dictionary for names and jargon, and snippets. With Auto-add to Dictionary on, it watches the text box it pasted into and adds words whose spelling you change, and Command Mode edits text by voice.
Handy keeps to one job. Its configurable shortcut records while held and stops on release, or toggles on a tap, with hold-only and toggle-only modes as well. It pastes the transcription into the focused app and restores your previous clipboard after a delay; a debug menu (Cmd+Shift+D or Ctrl+Shift+D) holds the paste-timing controls. Command-line flags such as handy --toggle-transcription and handy --cancel control a running copy, and a community Raycast extension on the Mac can start recording, browse transcript history, manage the dictionary, and switch models and languages.
5. Shell prompts, SSH sessions, and Wayland desktops
Wispr’s terminal article lists direct paste in Command Prompt, PowerShell, Windows Terminal, and WezTerm on Windows; Terminal.app and iTerm2 on a Mac; and Warp, Ghostty, kitty, Alacritty, Hyper, and VS Code terminals on both. In WSL windows, Linux virtual machines, and remote SSH sessions, Flow still transcribes but cannot paste, so you recover the text with Paste last transcript: Ctrl+Cmd+V on a Mac, Alt+Shift+Z on Windows. Wispr also warns that context-aware formatting can change shell commands, and that Flow must run as administrator when the target app does.
Handy’s README does not list terminals, so test it in the one you use with a harmless sentence first. Linux is where Handy says the most. It recommends xdotool on X11, wtype or dotool on Wayland (dotool needs your user in the input group), and ydotool on Ubuntu 26.04, where wtype does not work. On Wayland, global shortcuts must be set in the desktop environment, and the README gives the steps for GNOME, KDE Plasma, Sway, i3, and Hyprland, each binding handy --toggle-transcription. Flow has no Linux app to compare.
- Flow in WSL or over SSH: dictate, then Alt+Shift+Z or Ctrl+Cmd+V.
- Handy on Wayland: bind handy --toggle-transcription to a desktop shortcut.
- With either tool, dictate prose and agent prompts; type exact shell syntax yourself.
6. A third option, labelled as ours: IraVoice is built on Handy
Disclosure first: IraVoice publishes this page and is a commercial app built on Handy’s MIT-licensed code. IraVoice 1.2.2 keeps Handy’s on-device approach, recognizing speech with NVIDIA Parakeet; the default Parakeet Unified EN model, about 700 MB, downloads once and then works offline. It supports fewer systems than Handy: macOS 13+ on Apple silicon, Windows 10 or 11 (64-bit), and Linux (beta). That means no Intel Macs, x64-only Windows, and x64-only Linux (an AppImage, .deb or .rpm).
IraVoice’s own additions are aimed at people who dictate into code tools. Hold Right Command on a Mac or Right Ctrl on Windows, release to type, and press Esc to cancel. In terminals, editors, and coding agents on Mac and Windows it spells names from the Git project you are working in, so “use effect” becomes useEffect, and it learns words you correct. IraVoice Local costs $79 once for one computer, $109 for two, or $149 for three, after a 3-day free trial with no email, card, or account. If Handy already does what you need, it is free, and you should keep using it.