Modern Creator
Why Modern Creator?
Better Stack · YouTube

Open-Source Dictation Is Here — Goodbye Subscriptions

FluidVoice, a free open-source Mac app, runs speech-to-text and cleanup entirely on-device — going head-to-head with paid subscription tools like Wispr Flow and Superwhisper.

Posted
1 months ago
Duration
Format
Review
educational
Views
14.3K
587 likes
Big Idea

The argument in one line.

FluidVoice, a free open-source Mac app, runs both speech recognition and text cleanup entirely on-device, matching or beating paid subscription dictation tools like Wispr Flow without a monthly fee or cloud upload.

Who This Is For

Read if. Skip if.

READ IF YOU ARE…
  • A Mac developer who dictates code comments, commit messages, or docs and doesn't want a monthly subscription for it.
  • Anyone on Apple silicon who's cautious about voice audio or code leaving their machine and hitting a cloud server.
  • A Wispr Flow or Superwhisper user curious whether a free local alternative can match the paid cleanup quality.
SKIP IF…
  • You're on Windows or iOS — FluidVoice is Mac-only right now, with those platforms only on a wait list.
  • You dictate primarily in a language other than English — you may need extra tuning to get accurate results.
TL;DR

The full version, fast.

FluidVoice is a free, open-source Mac dictation app that runs entirely on-device: a local model called Parakeet transcribes speech, and a second local model, Fluid Intelligence, cleans up capitalization, punctuation, and structure before dropping the result into whatever app has focus. In a live demo it produced correctly formatted, comment-ready text with no lag, and it runs roughly four times faster than cloud competitors like Wispr Flow and Superwhisper because there's no network round-trip. The tradeoffs: it's Mac-only (Windows/iOS are wait-listed), the cleanup model is a ~3.5GB download, and it's slower on Intel Macs. The recommended move is to run it as your everyday Mac tool and keep a cross-platform tool around for when you're not on a Mac.

Free for members

Chat with this breakdown — free.

Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.

Create a free account →
Chapters

Where the time goes.

00:00 – 00:27

01 · Why Mac dictation isn't good enough

Apple's built-in dictation is fine for texts but fails at capitalization, punctuation, and structure for anything someone will actually read.

00:27 – 00:55

02 · FluidVoice: a free, open-source alternative

Introduces FluidVoice — free, open source, and fully on-device, addressing both the privacy and cost objections at once.

00:55 – 01:21

03 · How it works: Parakeet + local AI

Parakeet transcribes speech locally; Fluid Intelligence cleans it up; one hotkey drops the result into whatever app has focus.

01:21 – 02:12

04 · Live demo in VS Code

One-line install, mic/accessibility permissions, then a live hotkey dictation into a Claude Code terminal that produces clean, comment-formatted text with no visible lag.

02:12 – 02:43

05 · FluidVoice vs Apple Dictation, Wispr Flow, Superwhisper

FluidVoice runs about 4x faster than the cloud tools; Wispr Flow and Superwhisper are accurate but paid and cloud-dependent.

02:43 – 03:12

06 · Benefits: free, private, fast

Near-instant on Apple silicon; the 'weakest' option on paper turns out to be the only free, open-source, locally-formatted one.

03:12 – 03:43

07 · Limitations: Mac-only, model size, Intel support

Windows/iOS are wait-listed, the cleanup model is a ~3.5GB download, Intel runs slower, and non-English dictation may need tuning.

03:43 – 03:55

08 · Is it worth it for Mac developers?

A conditional verdict: worth it specifically if avoiding another subscription matters to you.

03:55 – 04:19

09 · Best way to use it (dual-tool workflow)

Run FluidVoice as the everyday Mac tool, keep a cross-platform tool around for when you're not on a Mac.

04:19 – 04:33

10 · GitHub link and Homebrew install command

Closing CTA: GitHub repo and brew install command in the description, subscribe ask, silent Better Stack end-card bumper.

Atomic Insights

Lines worth screenshotting.

  • FluidVoice runs entirely on-device: a local model called Parakeet transcribes speech, and a second local model, Fluid Intelligence, cleans up punctuation and structure — no audio ever reaches a server.
  • The free, open-source dictation tool beats the paid ones on the two things developers actually complain about: privacy and a recurring bill.
  • FluidVoice runs roughly four times faster than competing dictation tools because there's no cloud round-trip to wait on.
  • The cleanup model alone is about a 3.5 GB download, so setup takes longer than a typical one-line install even though the install command itself is a single line.
  • On Apple silicon (M-series chips), the local models run fast enough that dictation feels instantaneous; on Intel Macs it still works, just slower.
  • FluidVoice is Mac-only for now — Windows and iOS support exists only as a wait list, so Windows users still need a separate tool.
  • Apple's free built-in Mac dictation is fine for a text message but doesn't capitalize, punctuate, or structure text well enough for a commit message or a document someone will actually read.
  • A single hotkey captures speech and drops cleaned-up text directly into whatever app has focus — a terminal, Slack, Notes, or an AI coding tool — without picking a target app first.
Takeaway

Two small local models can replace one paid cloud subscription

WHAT TO LEARN

FluidVoice shows that chaining two small on-device models — one for transcription, one for cleanup — can match paid cloud dictation tools on quality while removing the monthly fee and the privacy risk.

01Why Mac dictation isn't good enough
  • Apple's built-in Mac dictation is fine for a text message, but it doesn't capitalize, punctuate, or structure a sentence well enough for a commit message or a document someone will actually read.
  • The two real objections developers have to dictation tools are privacy (recordings leaving the machine) and cost (another monthly subscription) — not accuracy.
02FluidVoice: a free, open-source alternative
  • FluidVoice is free and open source, and every part of the speech-to-text process happens on the Mac itself — nothing is uploaded to a server.
  • Running entirely on-device removes both developer objections at once: nothing leaves the machine, and there's no recurring bill.
03How it works: Parakeet + local AI
  • A local model called Parakeet transcribes raw speech, then a second local model, Fluid Intelligence, acts as an editor — fixing capitalization, punctuation, and structure before the text lands.
  • One hotkey drops the cleaned-up text into whatever app has focus — a terminal, Slack, Mail, or Notes — so there's no per-app integration to set up.
04Live demo in VS Code
  • Setup is a one-line install followed by granting mic and accessibility permissions — the same ritual every Mac app requires, not extra friction specific to this tool.
  • In a live demo, holding the hotkey and speaking a full sentence into a code editor produced correctly capitalized, punctuated text formatted like an actual code comment with no visible lag.
05FluidVoice vs Apple Dictation, Wispr Flow, Superwhisper
  • FluidVoice runs about four times faster than the cloud competitors because there's no network round-trip to wait on.
  • Wispr Flow and Superwhisper are accurate paid tools, but they still cost a subscription and still send audio to the cloud — for many developers that second point is the dealbreaker before price even comes up.
06Benefits: free, private, fast
  • On Apple silicon, the local models run fast enough that the dictation is effectively instant.
  • The tool expected to be the weakest option (free, open source) turned out to be the only one doing smart local formatting without a subscription or a cloud trip.
07Limitations: Mac-only, model size, Intel support
  • FluidVoice is Mac-only right now — Windows and iOS exist only as a wait list — and the cleanup model alone is about a 3.5 GB download, so setup takes longer than the one-line install suggests.
  • It works on Intel Macs, just slower than Apple silicon, and dictating in a language other than English may need extra tuning to get accurate results.
08Is it worth it for Mac developers?
  • The honest verdict is conditional, not a blanket yes: it's worth it specifically if avoiding another monthly subscription matters to you.
09Best way to use it (dual-tool workflow)
  • The recommended setup is to run FluidVoice as the everyday tool on Mac while keeping a cross-platform tool around for when you're not on a Mac — combining free local performance with platform coverage.
Glossary

Terms worth knowing.

Parakeet
The local speech-recognition model FluidVoice uses to convert spoken audio into raw text entirely on-device.
Fluid Intelligence
FluidVoice's second local model, which edits Parakeet's raw transcript — fixing capitalization, punctuation, and structure before it's inserted.
Wispr Flow
A paid, subscription-based cloud dictation tool that FluidVoice is positioned against in this video.
Superwhisper
Another paid dictation competitor mentioned as accurate but still cloud-dependent and subscription-priced.
Resources

Things they pointed at.

00:00toolApple's built-in Mac dictation
00:27toolFluidVoice ↗
00:06productWispr Flow
02:29productSuperwhisper
04:31productBetter Stack ↗
Quotables

Lines you could clip.

00:06
“You'd think the free thing we already have on Mac would be good enough, but it's not.”
relatable complaint that reframes the built-in tool as inadequate→ TikTok hook↗ Tweet quote
03:20
“This is actually four times faster than using other ones already out there.”
concrete, quotable performance claim→ IG reel cold open↗ Tweet quote
04:13
“On Apple silicon, it's fast enough that you can forget it's even thinking.”
tight, vivid one-liner on speed→ newsletter pull-quote↗ Tweet quote
04:02
“The free one that was supposed to be the weakest is the only one that's free open source and does the smart formatting right there on our machine.”
punchline that flips the expected hierarchy→ TikTok hook↗ Tweet quote
The Script

Word for word.

Read-along

Don't just watch it. Burn it in.

See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.

metaphor
Just to talk to your computer instead of typing it all out, you end up with another subscription to something like Whisper Flow. And, yeah, it's good. You'd think the free thing we already have on Mac would be good enough, but it's not.
This is Fluid Voice. It runs completely on your own machine and it's growing in popularity as a strong competitor to Whisper Flow and all the others.
Fluid Voice is a free open source Mac app that turns your voice into text, and every part of that happens on your own computer. Nothing ever gets shipped off to a server somewhere, which is a huge win. Two reasons this is good for us, and there are the exact two things devs complain about.
First, your voice and whatever you're dictating might be code or private, and this never leaves your Mac. Second, you're not paying another monthly bill to do it.
Here's how simple this is. A model called Parakeet listens and writes down what you say. Then a second local model called Fluid Intelligence acts like an editor sitting right next to you, fixing your capitalization, punctuation, and structure as you talk.
You hit one hotkey that you set, speak, and the cleaned up text drops into whatever your cursor is already at, into cursor clod, Slack, Mail's notes. It doesn't really care which one. Now this is super simple to spin up, so I'll keep the demo short too.
It's a one line install. You open it, give it mic and accessibility permissions. Every Mac app makes you do this, so you have to do it anyways.
And then you can choose a hotkey or use the one built in. So there's no real difference here. Now, watch this.
I hold the key and I just talk normally here. Can you check over my code and make sure it's set up with some good practices to prevent bugs? And it didn't just collapse.
Look at what it just did. It capitalized the line, punctuated it, kept my technical words spelled right, and formatted it like an actual comment. It didn't fix one thing.
No spinner, no waiting because there's no cloud to actually wait on here. So where does this actually sit next to everything else? Well, this is actually four times faster than using other ones already out there.
And on top of that, this is where it gets interesting. The dictation already on your Mac is free and for a text message, it's fine. For a commit message or a doc, someone actually gonna read that, it's probably not so great.
Whisperflow is the paid, so you'd expect it just wins here. But it's subscription, your audio goes to a cloud.
For a lot of us, that second part is a deal breaker before we even get to the price. Super Whisper and the rest are good tools. Still cost money.
They don't quite match the local cleanup Fluid Voice is actually doing. So the free one that was supposed to be the weakest is the only one that's free open source and does the smart formatting right there on our machine.
On Apple silicon, it's fast enough that you can forget it's even thinking. Now, before you run off and install this, it's great that it's free. Right?
There's no subscription. It's great that it's private. Our voice stays on our Mac.
On m series machines, it's fast and the formatting is actually pretty good. And it's open source and actively worked on, so problems are getting fixed. But on the flip side of things, it's Mac only right now.
IOS and Windows are on the wait list. So if you live on Windows, this isn't your tool just yet.
The editor model is about three and a half gigs, so it's a real download. It's not a quick one. It's best, like I said, on the silicone chips.
It does run on Intel, it's just slower. And if you dictate in a language other than English, you might have to tune it to get it just right. But you should know all that now, not after you spend ten minutes trying to download this the hard way.
You probably expect me to say it's worth it now, but I won't. For most Mac devs, yeah, okay, this might be worth giving a shot.
It's worth it if you don't want another monthly payment just to dictate things for you, which come on, who does want that? Use it for boring typing. Email, docs, code comments, Slack messages, heck, talk to Claude.
On a Mac m four pro, it's really nice because it's all local and it's quick. If you need Windows or iOS right now, sorry, but this can't help you. So here's what you could actually do.
Run it as your everyday tool in Mac and keep something cross platform around for when you're not on Mac. You get the best of both worlds and it costs you nothing if you're using Mac. The GitHub link and the brew command are both in the description below.
If you enjoy coding tips and tricks like this, be sure to subscribe to the BetterStack channel. We'll see you in another video.
The Hook

The bait, then the rug-pull.

Every Mac already has a dictation button — and it's not good enough for real work. This video tests the free, open-source app built to replace it, and the subscription tools built on top of it.

Frameworks

Named ideas worth stealing.

00:55model

FluidVoice's two-model pipeline

  1. Parakeet — local speech-to-text
  2. Fluid Intelligence — local cleanup & formatting

Speech runs through two on-device models in sequence: one converts audio to raw text, the second acts as an editor fixing capitalization, punctuation, and structure before the hotkey drops it into the active app.

Steal forany pitch comparing a single big cloud model against two small chained local models
CTA Breakdown

How they asked for the click.

VERBAL ASK
04:19link
“The GitHub link and the brew command are both in the description below. If you enjoy coding tips and tricks like this, be sure to subscribe to the BetterStack channel.”

single closing line pairs the tool's install link with a subscribe ask — no hard sell, no separate spoken sponsor read despite the channel's own Better Stack product appearing as a silent end-card bumper

MENTIONED ON CAMERA
00:27toolFluidVoice ↗
04:31productBetter Stack ↗
FROM THE DESCRIPTION
PRIMARY CTAWhere the creator wants you to go next.
OTHER LINKSAlso linked in the description.
Storyboard

Visual structure at a glance.

open
hookopen00:00
intro FluidVoice
promiseintro FluidVoice00:27
live demo
valuelive demo01:36
CTA
ctaCTA04:19
Frame Gallery

Visual moments.

open00:01Frame at 00:05 from Open-Source Dictation Is Here — Goodbye Subscriptions00:05Frame at 00:07 from Open-Source Dictation Is Here — Goodbye Subscriptions00:07Frame at 00:11 from Open-Source Dictation Is Here — Goodbye Subscriptions00:11Frame at 00:15 from Open-Source Dictation Is Here — Goodbye Subscriptions00:15Frame at 00:19 from Open-Source Dictation Is Here — Goodbye Subscriptions00:19
open00:01Frame at 00:05 from Open-Source Dictation Is Here — Goodbye Subscriptions00:05Frame at 00:07 from Open-Source Dictation Is Here — Goodbye Subscriptions00:07Frame at 00:11 from Open-Source Dictation Is Here — Goodbye Subscriptions00:11Frame at 00:15 from Open-Source Dictation Is Here — Goodbye Subscriptions00:15Frame at 00:19 from Open-Source Dictation Is Here — Goodbye Subscriptions00:19Frame at 00:23 from Open-Source Dictation Is Here — Goodbye Subscriptions00:23Frame at 00:25 from Open-Source Dictation Is Here — Goodbye Subscriptions00:25intro FluidVoice00:27Frame at 00:32 from Open-Source Dictation Is Here — Goodbye Subscriptions00:32Frame at 00:36 from Open-Source Dictation Is Here — Goodbye Subscriptions00:36Frame at 00:38 from Open-Source Dictation Is Here — Goodbye Subscriptions00:38Frame at 00:42 from Open-Source Dictation Is Here — Goodbye Subscriptions00:42Frame at 00:45 from Open-Source Dictation Is Here — Goodbye Subscriptions00:45Frame at 00:49 from Open-Source Dictation Is Here — Goodbye Subscriptions00:49Frame at 00:52 from Open-Source Dictation Is Here — Goodbye Subscriptions00:52Frame at 00:56 from Open-Source Dictation Is Here — Goodbye Subscriptions00:56Frame at 01:00 from Open-Source Dictation Is Here — Goodbye Subscriptions01:00Frame at 01:03 from Open-Source Dictation Is Here — Goodbye Subscriptions01:03Frame at 01:08 from Open-Source Dictation Is Here — Goodbye Subscriptions01:08Frame at 01:10 from Open-Source Dictation Is Here — Goodbye Subscriptions01:10Frame at 01:13 from Open-Source Dictation Is Here — Goodbye Subscriptions01:13Frame at 01:16 from Open-Source Dictation Is Here — Goodbye Subscriptions01:16Frame at 01:19 from Open-Source Dictation Is Here — Goodbye Subscriptions01:19Frame at 01:23 from Open-Source Dictation Is Here — Goodbye Subscriptions01:23Frame at 01:26 from Open-Source Dictation Is Here — Goodbye Subscriptions01:26Frame at 01:30 from Open-Source Dictation Is Here — Goodbye Subscriptions01:30Frame at 01:33 from Open-Source Dictation Is Here — Goodbye Subscriptions01:33Frame at 01:38 from Open-Source Dictation Is Here — Goodbye Subscriptions01:38Frame at 01:40 from Open-Source Dictation Is Here — Goodbye Subscriptions01:40Frame at 01:44 from Open-Source Dictation Is Here — Goodbye Subscriptions01:44live demo01:47Frame at 01:50 from Open-Source Dictation Is Here — Goodbye Subscriptions01:50Frame at 01:54 from Open-Source Dictation Is Here — Goodbye Subscriptions01:54Frame at 01:57 from Open-Source Dictation Is Here — Goodbye Subscriptions01:57Frame at 02:01 from Open-Source Dictation Is Here — Goodbye Subscriptions02:01Frame at 02:05 from Open-Source Dictation Is Here — Goodbye Subscriptions02:05Frame at 02:07 from Open-Source Dictation Is Here — Goodbye Subscriptions02:07Frame at 02:11 from Open-Source Dictation Is Here — Goodbye Subscriptions02:11Frame at 02:14 from Open-Source Dictation Is Here — Goodbye Subscriptions02:14Frame at 02:18 from Open-Source Dictation Is Here — Goodbye Subscriptions02:18Frame at 02:21 from Open-Source Dictation Is Here — Goodbye Subscriptions02:21Frame at 02:26 from Open-Source Dictation Is Here — Goodbye Subscriptions02:26Frame at 02:28 from Open-Source Dictation Is Here — Goodbye Subscriptions02:28Frame at 02:30 from Open-Source Dictation Is Here — Goodbye Subscriptions02:30Frame at 02:35 from Open-Source Dictation Is Here — Goodbye Subscriptions02:35Frame at 02:37 from Open-Source Dictation Is Here — Goodbye Subscriptions02:37Frame at 02:42 from Open-Source Dictation Is Here — Goodbye Subscriptions02:42Frame at 02:45 from Open-Source Dictation Is Here — Goodbye Subscriptions02:45Frame at 02:48 from Open-Source Dictation Is Here — Goodbye Subscriptions02:48Frame at 02:52 from Open-Source Dictation Is Here — Goodbye Subscriptions02:52Frame at 02:55 from Open-Source Dictation Is Here — Goodbye Subscriptions02:55Frame at 02:59 from Open-Source Dictation Is Here — Goodbye Subscriptions02:59Frame at 03:02 from Open-Source Dictation Is Here — Goodbye Subscriptions03:02Frame at 03:05 from Open-Source Dictation Is Here — Goodbye Subscriptions03:05Frame at 03:09 from Open-Source Dictation Is Here — Goodbye Subscriptions03:09Frame at 03:12 from Open-Source Dictation Is Here — Goodbye Subscriptions03:12Frame at 03:16 from Open-Source Dictation Is Here — Goodbye Subscriptions03:16Frame at 03:20 from Open-Source Dictation Is Here — Goodbye Subscriptions03:20Frame at 03:23 from Open-Source Dictation Is Here — Goodbye Subscriptions03:23Frame at 03:24 from Open-Source Dictation Is Here — Goodbye Subscriptions03:24Frame at 03:29 from Open-Source Dictation Is Here — Goodbye Subscriptions03:29Frame at 03:33 from Open-Source Dictation Is Here — Goodbye Subscriptions03:33Frame at 03:36 from Open-Source Dictation Is Here — Goodbye Subscriptions03:36Frame at 03:40 from Open-Source Dictation Is Here — Goodbye Subscriptions03:40Frame at 03:43 from Open-Source Dictation Is Here — Goodbye Subscriptions03:43Frame at 03:46 from Open-Source Dictation Is Here — Goodbye Subscriptions03:46Frame at 03:50 from Open-Source Dictation Is Here — Goodbye Subscriptions03:50Frame at 03:53 from Open-Source Dictation Is Here — Goodbye Subscriptions03:53Frame at 03:56 from Open-Source Dictation Is Here — Goodbye Subscriptions03:56Frame at 04:00 from Open-Source Dictation Is Here — Goodbye Subscriptions04:00Frame at 04:03 from Open-Source Dictation Is Here — Goodbye Subscriptions04:03Frame at 04:07 from Open-Source Dictation Is Here — Goodbye Subscriptions04:07Frame at 04:10 from Open-Source Dictation Is Here — Goodbye Subscriptions04:10Frame at 04:14 from Open-Source Dictation Is Here — Goodbye Subscriptions04:14Frame at 04:17 from Open-Source Dictation Is Here — Goodbye Subscriptions04:17CTA04:21Frame at 04:24 from Open-Source Dictation Is Here — Goodbye Subscriptions04:24Frame at 04:27 from Open-Source Dictation Is Here — Goodbye Subscriptions04:27Frame at 04:31 from Open-Source Dictation Is Here — Goodbye Subscriptions04:31
One-click upgrade to your Google

Get more breakdowns in your search results

Add Modern Creator as a preferred source and Google shows you more of our breakdowns in Search, Top Stories, and AI Overviews. It only changes what you see, and you can undo it in your Google settings anytime.

Add to Preferred SourcesOpens your Google source preferences with us pre-loaded. Tick the box and you're done.
Watch next

More from this channel + related breakdowns.

11:25
Matt Pocock · Tutorial

I Open-Sourced My Own AFK Software Factory

Matt Pocock built and open-sourced Sandcastle, a TypeScript library that runs Claude Code and other coding agents inside sandboxes to plan, implement, review, and merge whole GitHub issues without a human clicking approve.

April 30th