FluidVoice vs Handy: One Is Open Source. The Other Really Is.

I spent three weeks with FluidVoice as my daily dictation app. Handy is the one people keep naming as the alternative. Both are free, both live on GitHub, both keep your audio on your machine — and on the one thing they both claim, they don't actually match.

Handy goes all the way. MIT-licensed, around 30,500 stars, builds for macOS, Windows and Linux, and — this is the bit that counts — there's no cloud mode anywhere in it. Nothing to misconfigure. There's also no AI cleanup, because that would mean shipping a model you'd have to take on trust.

FluidVoice makes the practical call instead. GPLv3 app, on-device speech models, and a closed model called Fluid-1 that runs on your Mac and makes the output a lot nicer to read. The developers don't hide the arrangement: keeping Fluid Intelligence private is what pays for the free app.

So it isn't really a privacy question — both are local. It's whether you'd rather have cleaner text you can't check, or rougher text you can.

All figures retrieved from each project's repository and documentation on August 28, 2026.

Nobody here gives you everything. Handy gives up output quality so you can read the whole stack. FluidVoice gives up one closed component so your text arrives usable. Paid apps give up source access so you get more platforms and someone to email.

Worth asking which of those you'd miss. In practice people notice the thing they gave up, not the thing they got.

FluidVoice is a free macOS dictation app from Altic, public since September 2025 and moving fast — 11,006 GitHub stars, 762 forks, and v1.6.9 shipped on August 18, 2026. It is licensed GPLv3 as of February 23, 2026, having previously been Apache 2.0.

The speech layer is entirely local, with an unusually wide model lineup: Nemotron Speech 3.5, Parakeet Flash and TDT v2 and v3, Cohere Transcribe, Apple Speech, and Whisper. On top of that sits Fluid Intelligence, which runs the custom-trained Fluid-1 model — roughly a 3.5 GB optional download — to handle smart formatting, capitalisation, and tone. It needs no API key and works offline.

Fluid-1 is also the app's one closed component, which the README states outright: it is "a separate, privately maintained local AI runtime," kept private "so we can sustainably offer the core dictation experience for free." We unpack what that does and doesn't mean for privacy in is FluidVoice safe.

Handy is a free MIT-licensed dictation app built in Rust and Tauri, and it's the biggest open-source project in this category — around 30,500 stars and 2,700 forks as of August 28, 2026. It runs on macOS (Intel and Apple Silicon), Windows and Linux, the only app here that covers all three.

Most of what defines it is what it leaves out. No cloud mode, no account, no telemetry, no AI rewriting. Local models include Whisper in several sizes, Parakeet V2 and V3, Moonshine and Cohere Transcribe, with Silero voice activity detection to skip the silence. Developers get CLI flags and a Raycast extension.

What comes back is near-verbatim text with barely any auto-punctuation, usually two to five seconds after you stop talking on Whisper models. Some people hit occasional first-word clipping. Our full Handy review covers living with it, and is Handy safe covers how it's built.

Handy takes five of those six. The one it loses — output quality — is the one you notice on every single dictation, which is why FluidVoice keeps getting recommended anyway.

Two things settle it for most people. Platform: FluidVoice needs macOS 15 Sequoia or later, anything past Whisper needs Apple Silicon, and the Windows build is a 0.0.9 pre-release with an unsigned installer. Handy ships stable builds for Intel Macs, Windows and Linux. And the cleanup gap is bigger than it sounds — near-verbatim output means you're back to adding the punctuation and paragraph breaks yourself.

Neither app has a Product Hunt or App Store rating worth quoting, so GitHub adoption is the best signal available — and it isn't close. Handy sits at roughly 30,500 stars and 2,700 forks; FluidVoice at 11,006 and 762. Some of that is just age: Handy has been around longer, and FluidVoice's repository only opened in September 2025.

FluidVoice was also picked as the top macOS option in Adam Jones's independent test of 21 Wispr Flow alternatives, which is a genuine outside vote for the output quality.

Set against that, three weeks of using it daily turned up a run of problems: my microphone handling broke more than once, switching to AirPods froze dictation, and one update mid-test started clipping the first word of every sentence until I rolled back. The cleanup layer occasionally summarised instead of tidying. Handy's problems are duller — first-word clipping and the wait after you stop talking.

Roughly: FluidVoice tries more and breaks in more interesting ways. Handy does less and does it the same way every time.

Go with FluidVoice if you're on Apple Silicon running macOS 15 or later and you want your dictation to land already formatted without pasting an API key anywhere. Turn analytics off in the first minute. You're taking Fluid-1 on trust rather than reading it — that's the deal.

Go with Handy if you're on Windows, Linux, an Intel Mac or macOS 14 and earlier, where FluidVoice either won't run or is still a 0.0.9 pre-release. Same answer if one closed component rules an app out for you. You'll be punctuating your own text and waiting a few seconds for it.

Look at Voibe if neither of those works — you need Mac and Windows on one licence, you want the cleanup, and you want someone to chase when it breaks. It runs on Mac and Windows: the fully on-device Whisper mode is Mac-only (Apple Silicon), while Intel Macs and Windows use Voibe's zero-retention private cloud — self-hosted open-source models where audio is never stored, sold, or used to train AI. $7.50/mo, $59/yr, or $149 lifetime — Voibe for Windows covers the native Windows app. It's closed source, and that's what you're giving up for the rest of it.

Handy is the better open-source app. FluidVoice is the better dictation app. If "open source" is what you were after, Handy is what you meant — nothing closed inside it, and it runs on every desktop platform. If you're on a recent Apple Silicon Mac and you want your words to arrive punctuated, FluidVoice does that for free, and the part you're trusting never leaves your machine.

The one bad outcome is picking FluidVoice because it's open source without realising Fluid-1 isn't. Once you know that, either choice holds up.

• Is FluidVoice Safe? — the full breakdown of what's open, what's closed, and where each piece runs.
• FluidVoice Alternatives — what to run on Windows, Intel Macs, and macOS 14 or earlier.
• FluidVoice Review — three weeks of daily dictation, including what broke.
• Handy Review — the free MIT-licensed offline app in day-to-day use.
• OpenWhispr vs Handy — the same open-source question with a cloud tier attached.
• Best Open-Source Wispr Flow Alternatives — the wider field, ranked.
• All Dictation App Comparisons — the hub for every head-to-head on the site.