LumeVoice vs Handy: Free and Open Source, or Cleanup Everywhere?

LumeVoice vs Handy comes down to two questions: does your audio have to stay on your computer, and do you want finished text without setting anything up? Handy transcribes on your own Mac, Windows or Linux machine, costs nothing, and its code is open for anyone to read. LumeVoice transcribes in the cloud, strips filler words, applies your self-corrections, and also runs on iPhone and Android.

A quick disclosure: LumeVoice is our product, and we haven't used Handy, so nothing below is a hands-on test. What we say about Handy comes from its website, documentation and GitHub repository, all checked on October 20, 2026, when version 0.9.8 was on its download page. Handy changes quickly, so check its docs for the latest. Our other head-to-heads are on the comparisons page.

Who should pick which#

Pick Handy if you:

• want dictation that costs nothing, now or later
• need your audio to stay on your computer, or work without a connection
• use Linux
• like open-source code you can read, change or build yourself
• want to choose your own speech model, from tiny English-only models to Whisper for 99+ languages
• only dictate on a computer

Pick LumeVoice if you:

• want text that's ready to send, with fillers, stutters and self-corrections ("five, actually six" becomes "six") handled for you
• also dictate on an iPhone or Android phone
• would rather not download models, choose between them or set up an AI provider
• want snippets, writing styles and an output language (speak English, get Spanish) built in

LumeVoice vs Handy at a glance#

Handy details as of October 2026, from handy.computer, its docs and its GitHub repository.

Price and 3-year cost#

Handy is free. There's no paid plan, trial or license key; its site says accessibility tooling belongs in everyone's hands, "not behind a paywall", and the project runs on sponsors and donations (Handy). The one cost you can opt into is post-processing with a cloud AI provider such as OpenAI, Anthropic, Groq or OpenRouter, which bills you for what you use (post-processing).

Here's what each option costs over three years:

On price, Handy wins outright. LumeVoice only makes sense if it saves you enough editing, or covers enough of your devices, to be worth paying for.

LumeVoice starts every new account with 3,000 free words, no card required, and Pro Monthly and Pro Annual begin with a 7-day free trial the first time you upgrade in the app (card required). Students and teachers get 50% off, and anyone who finds typing hard because of a disability, injury or health condition gets 50% off, with no proof needed. Codes are entered at checkout on Mac or Windows. See LumeVoice pricing, student pricing and the accessibility discount.

Platforms: where each app runs#

Mac. Handy has separate downloads for Apple silicon and Intel Macs, plus a community-maintained Homebrew cask (download). It asks for Microphone and Accessibility permissions, and its default shortcut is Option + Space (getting started). LumeVoice lives in the menu bar: hold fn (Globe), speak and let go. It runs on macOS 11.5 or later, on Apple silicon and Intel Macs. More on LumeVoice for Mac.

Windows. Both. Handy has .exe and .msi installers for x64 and ARM PCs, plus a community-maintained winget package. Its FAQ notes that it can't paste into apps running as administrator unless Handy runs as administrator too. LumeVoice sits in the system tray: hold Ctrl + Win and talk, tap Space while holding to go hands-free, or press Esc to cancel. It installs per user, with no admin rights needed. See LumeVoice for Windows.

Linux. Only Handy, on x64 and ARM, as an AppImage, .deb or .rpm. Its README describes Wayland support as limited and lists the typing tools each setup needs: xdotool on X11, and wtype, dotool or ydotool on Wayland. Its FAQ adds that the AppImage needs GLIBC 2.38 or later, which rules out Ubuntu 22.04. LumeVoice has no Linux app.

iPhone and Android. Only LumeVoice. On iPhone it's the Lume keyboard (iOS 17 or later): switch to it in any app, tap Start Lume and talk. On Android it's a floating bubble over your keyboard: tap it, talk, tap the check mark, and you keep your usual keyboard. Handy has no phone apps. See LumeVoice for iPhone and LumeVoice for Android.

Features: models, cleanup and post-processing#

Both apps type into whatever app has your cursor and keep a list of words for names and jargon: custom words in Handy, the custom dictionary in LumeVoice. Here's where they differ.

Speech models. Handy is built around choice (models). Parakeet V3 is its recommended default: quick on an ordinary processor, with 25 European languages detected automatically. Whisper models (Small, Medium, Large and Turbo) cover 99+ languages and use your graphics chip through Metal on a Mac or Vulkan on Windows and Linux. Cohere Transcribe aims at the best English, and small Moonshine models suit slower machines. Downloads range from about 30 MB to 1.8 GB. LumeVoice makes the choice for you: speech is transcribed in the cloud by Whisper large-v3-turbo running on Groq, with Mistral's Voxtral as a fallback, and a language model then does the cleanup.

Cleanup. In LumeVoice, cleanup is the point: it removes filler words and stutters, applies self-corrections ("let's meet at five, actually six" becomes "Let's meet at six."), turns spoken lists into numbered or bulleted lists, and fixes grammar and spelling, with nothing to set up. Handy's README sums up its approach as one tool with one job: put what you say into a text box. Its post-processing feature, marked alpha and switched on under Settings, Advanced, Experimental Features, sends the transcript to an AI model with a prompt you write (post-processing). You can use OpenAI, Anthropic, Groq, OpenRouter, Cerebras or Z.AI with your own API key, Apple Intelligence on an Apple silicon Mac running macOS 26, or a model on your own machine through any OpenAI-compatible server such as LM Studio. It runs from its own shortcut, so a normal dictation stays plain.

Quirks to know about. Handy's FAQ is candid: Parakeet writes numbers as words, Whisper sometimes adds a stray "thank you" after silence, recordings longer than three to five minutes may fail, and it transcribes your microphone only, not system audio (FAQ).

Extras. LumeVoice adds snippets (say a short cue, get your address or sign-off), writing styles, an output-language setting, a pill that offers a Copy button when there's no text field, and audio and video file transcription on the desktop. Handy adds a command-line interface, a community Raycast extension and an interface in 18+ languages, and most of its Whisper models can translate speech into English.

Privacy: where your voice goes#

• Handy: its FAQ says all speech-to-text transcription happens on your device, and that the app only reaches the network to download models and check for updates. If you turn on post-processing with a cloud provider, your transcripts go to that provider's API; Handy lists Apple Intelligence and your own local server as options that stay on your machine. Because the code is open, anyone can check how it behaves.
• LumeVoice: your audio is sent for processing, isn't stored after processing, and is never used to train models. Your dictation history stays on your device. It needs an internet connection.

The real difference: with Handy, nothing you say has to leave your computer, which LumeVoice can't offer. If your work requires that, choose Handy; our guide to dictation for legal and medical work covers what to check.

Where Handy is the better choice#

• Price: free, with nothing to unlock.
• Privacy: transcription on your own computer, with no connection needed once a model is downloaded.
• Open source under the MIT license: read it, change it, build it.
• Linux, on x64 and ARM.
• Model choice, from tiny English-only models to Whisper Large, plus custom Whisper models.
• Your own AI cleanup, if you want it, with any OpenAI-compatible model, including one on your own machine.

Where LumeVoice is the better choice#

• iPhone and Android, plus Mac and Windows, on one plan.
• Cleanup by default: filler words, stutters and self-corrections handled without prompts, keys or experimental settings.
• No models to manage: nothing to download, compare or keep up to date.
• Snippets, writing styles and an output language, built in.
• Published discounts: 50% off for students and teachers, and 50% off for anyone who finds typing hard.

Using both, or switching from Handy#

1. Keep Handy for material that must stay on your computer, or for your Linux machine.
2. Copy your custom words into the LumeVoice custom dictionary on Mac or Windows, and turn any post-processing prompt you relied on into a snippet or a writing style where one fits.
3. Give each app its own shortcut. LumeVoice uses fn on a Mac and Ctrl + Win on Windows by default, and Handy uses Option + Space on a Mac, so check nothing overlaps if you've changed them.
4. Use your free words on real work for a few days, on your computer and your phone.
5. Ask for your discount before you buy. Email hello@lumevoice.com for the student or accessibility code, then upgrade on Mac or Windows.

For more, see the best Handy alternatives, LumeVoice vs VoiceInk, LumeVoice vs Spokenly and our roundup of open-source Whisper apps.

The bottom line#

Handy is hard to argue with on value: it's free, open source and keeps your voice on your own computer, on Mac, Windows and Linux. LumeVoice is for people who'd pay to skip the setup and the editing, want the same app on their phone, and are fine with the cloud. If it has to be free, work without internet or run on Linux, choose Handy. If not, dictate a week of real writing in both and keep the one you edit less.