You've got a hotkey under your finger, a half-written email, and a sentence you'd rather say out loud than type. Then you remember the dictation tool you tried last year sent every word to a server somewhere, and you quietly close the tab. That's the itch TypeWhisper scratches.
This article tells you what TypeWhisper actually is, how it works on a Mac, what's free and what costs money (with the current numbers), where it shines, where it's a hassle, and how it stacks up against other voice tools, including GhostWriter. I'll give you a straight recommendation at the end.
What is TypeWhisper?
TypeWhisper is a free, open source dictation and transcription app that does speech recognition on your own device. It runs on macOS, Windows and iOS. According to the TypeWhisper website and its GitHub organization, all three platforms were listed as available as of October 2026, and core dictation is free, GPLv3-licensed, and designed to work without creating an account.
The basic loop is simple. You trigger dictation with a global hotkey, talk, and the transcribed text is inserted straight into the active application. No copy-paste dance. It's a speech-to-text tool only, so don't expect voice cloning or text-to-speech.
What makes it unusual is the engine choice. Rather than locking you into one model, TypeWhisper lets you pick between several local engines (WhisperKit, Parakeet TDT, Apple Speech) and optional cloud providers such as Groq, OpenAI and Gemini. Local is the default, and you have to switch cloud on yourself, one provider at a time.
Why people noticed it
Tom's Guide ran a piece on it describing a writer who was tired of expensive, demanding transcription software for interviews. He installed the WhisperKit model locally on a MacBook Air, settled on the Large v3 model, and came away impressed that it was free for non-commercial use and ran entirely on his machine. That's the pitch in a sentence: good transcription, no subscription required, nothing leaving your laptop.
I think that's a real selling point. Wispr Flow is quick, and so is built-in macOS dictation, but neither lets you say with a straight face, "my audio never left this computer." TypeWhisper does, as long as you stay on a local engine.
How TypeWhisper works on a Mac
On a Mac, TypeWhisper sits in the menu bar, listens when you hit a hotkey, and types the result into the focused app. The website lists macOS 14 or later as the requirement. The stable release as of October 2026 is version 1.7.1, shipped on October 6, 2026, per the TypeWhisper changelog.
Here's the typical setup flow:
- Download and install: grab the app from the TypeWhisper site or GitHub and drop it in Applications.
- Grant permissions: macOS will ask for microphone access and accessibility access (the latter is how any dictation tool types into other apps).
- Pick an engine: choose a local model. Smaller models are quick; larger ones are more accurate but heavier.
- Set your hotkey: pick something you can hit without thinking, ideally a single modifier or a function key.
- Dictate: hold or toggle the hotkey, speak, and watch the text appear.
Budget about ten minutes for all of that, most of it spent waiting for a model to download. After that, it's a menu bar icon you forget about.
Choosing a model
This is where most of the fiddling happens. Bigger Whisper-family models, like Large v3, give you better accuracy with accents, jargon and names, but they take more memory and a bit longer to process. On a recent Apple Silicon Mac, that's usually fine. On an older Intel machine, you'll feel it.
My honest take: start with a mid-size model, test it on your real voice and your real vocabulary for a day, and only then jump to the largest one. The Large v3 download alone runs to a few gigabytes, and if you grab it first and your Mac chugs, you'll blame the app when the real culprit is the model you picked.
Local versus cloud engines
Local engines keep audio on your machine. Cloud providers can be faster or more accurate for certain languages, but you're trusting a third party with your voice. TypeWhisper makes that a choice rather than a default, which is the right way round.
| Engine type | Examples | Privacy | Best for |
|---|---|---|---|
| Local | WhisperKit, Parakeet TDT, Apple Speech | Audio stays on device | Everyday dictation, sensitive text |
| Cloud (optional) | Groq, OpenAI, Gemini | Audio sent to provider | Hard languages, speed on weak hardware |
What's new: recent releases
The changelog shows a project that ships often and moves fast. A few recent items worth knowing, all from the TypeWhisper changelog and GitHub:
- macOS 1.7.1 (October 6, 2026): added configurable filler-word removal and English spelling variants (so you can choose US or UK spellings), and fixed a system-wide input freeze. That freeze fix matters. An app that can lock up your typing is an app you stop trusting.
- macOS 1.6 (August 2026): added backup and restore, calendar-aware meeting capture, and integrations including Soniox live transcription and Cohere Transcribe.
- macOS 1.8 (in progress): daily builds were already preparing it by October 13, 2026, with paired-device transcription over a local network and better meeting recordings. These are pre-release builds, so don't rely on them for important work.
The Windows version moves on its own track. Windows 1.0.9 is listed as stable, while 1.1 release candidates and 1.2 daily builds were appearing in October 2026. If you're on Windows, expect a slightly rougher ride than on Mac.
The iOS app
The iOS app launched on the App Store in August 2026. According to the TypeWhisper listing on the App Store, it needs iOS 18 or later and supports dictation in other apps through a keyboard, on-device models, optional cloud providers, and audio or video transcription.
A keyboard is the only way iOS lets third parties dictate into other apps, so TypeWhisper had no real alternative. Is it good enough? For short messages and quick notes, yes. You tap the globe key, switch keyboards, speak, and the text drops in. For anything longer than a paragraph, I'd still reach for a Mac. The switch is slower than a hotkey, and it never stops feeling like a workaround.
TypeWhisper pricing
Core dictation is free, and paid plans start at €5 per month. Per the TypeWhisper pricing page, the commercial tiers look like this:
| Plan | Price | Who it's for |
|---|---|---|
| Free | €0 | Individuals using core dictation, non-commercial |
| Individual | €5 per month | One person who wants the extras |
| Team | €19 per month | Up to 10 devices |
| Enterprise | €99 per month | Unlimited devices |
Paid features include synced dictionaries and snippets, learning from your corrections, and commercial licensing terms. That last one is easy to miss. The software is GPLv3, and Tom's Guide notes it's free for non-commercial use. If you're using it for paid work inside a company, check the licensing terms rather than assuming free means free for everything.
Is the paid tier worth it?
If you dictate lightly, no. The free version does the job. If you dictate all day and hate re-teaching the app the same product names and surnames, the synced dictionary and correction-learning features start to earn their €5. For a team, the per-device cap on the €19 plan is the thing to do the maths on before you commit. Ten devices sounds generous until you count laptops, desktops and phones per person.
Test any dictation tool on your own worst vocabulary, not a demo sentence. Names, acronyms and jargon are where tools fall apart.
Where TypeWhisper is genuinely good
Privacy, price and flexibility are the big three. Let me be specific.
- Local by default: you can dictate confidential drafts, client notes or interview audio without a cloud round trip.
- Open source: anyone can read the code. There's even a plugin SDK on GitHub for building transcription engines, text-to-speech providers, LLM providers, post-processors and custom actions. If you're a developer, that's a real draw.
- Engine choice: swap models as better ones appear. You're not stuck with whatever the vendor picked in 2024.
- Cross-platform: Mac, Windows and iOS under one roof is rare for a free tool.
- Fast iteration: the changelog cadence suggests bugs get fixed quickly.
The plugin angle deserves a note. Most dictation apps are closed boxes. TypeWhisper's SDK means a developer could wire in a new engine or a post-processing step that cleans up text for a specific purpose. Most people won't touch that. But it's why the project has a community around it.
Where it gets annoying
TypeWhisper asks more of you than a polished commercial app does. That's the trade-off for flexibility and a free price tag.
Setup is on you
Picking engines, downloading models and tuning a hotkey is fine if you enjoy tinkering. If you just want to talk and have clean text appear, the number of choices can feel like homework. A newcomer won't know whether to pick Parakeet or WhisperKit, and the app can't fully answer that for you.
Hardware matters
Large local models want a decent Mac. On an M-series MacBook Air, the Tom's Guide writer found it workable. On older hardware, you may be stuck with smaller, less accurate models, or pushed toward cloud providers, which undoes the privacy advantage.
Rapid releases cut both ways
A project that ships fast also ships regressions. The system-wide input freeze fixed in 1.7.1 is a good example of the kind of bug that can bite when an app hooks into your whole system. Stick to stable releases, not the daily builds, unless you like living dangerously.
The platforms aren't equal
Windows is on a separate release stream with its own version numbers, and iOS is brand new. If you split your day across all three, expect uneven polish.
TypeWhisper versus other Mac dictation options
TypeWhisper wins on cost and privacy; paid tools usually win on out-of-the-box polish. Here's the landscape in short.
- Built-in macOS dictation: free and zero setup, but limited in formatting and tone. Our guide on how to turn on voice dictation on every device walks through it.
- Wispr Flow: a polished, cloud-first tool with a free tier that has limits. We've covered whether Wispr Flow is free and whether it's safe in separate pieces, and the privacy question is the main reason people look at local alternatives like TypeWhisper.
- GhostWriter: a macOS app with voice-powered AI that turns speech into clean, ready-to-paste text in any application, with punctuation and formatting handled automatically and tone adapted for contexts like emails or code editors.
Where GhostWriter fits
Full disclosure: this article lives on GhostWriter's site, so weigh my comparison accordingly. TypeWhisper gives you a toolbox: engines, models, plugins, and a licence to match. GhostWriter gives you a finished workflow. You speak, and what arrives is already punctuated, formatted, and shaped to the app you're typing into, whether that's an email or a code editor. If you enjoy choosing models and tinkering, TypeWhisper is great. If you'd rather not think about models at all and want the tone adjusted for you, GhostWriter is the shorter path.
That said, if your top priority is a free, fully open source, fully local setup, TypeWhisper is hard to beat. GhostWriter doesn't try to be that.
Who should use TypeWhisper
If you value privacy, like control, or can't justify a subscription, this is your tool. Specifically:
- Journalists and researchers: transcribing interviews locally keeps sources' voices off third-party servers.
- Developers: the open source code and plugin SDK make it hackable.
- Privacy-minded writers: drafting sensitive material without a cloud dependency.
- Budget-conscious users: free core dictation beats paying monthly for basic transcription.
Skip it, or at least be cautious, if you want zero configuration, you're on an older Mac, or you need a single polished experience across several devices on day one.
Using it for transcription, not just dictation
Dictation and transcription are cousins, not twins. Dictation is you talking into a text field in real time. Transcription is turning an existing recording, like an interview or a lecture, into text. TypeWhisper handles both, and the iOS app explicitly supports audio and video transcription. If your main need is turning old recordings into text, our rundown of free audio transcription options that actually work is a useful companion read.
Practical tips to get better results
Your habits matter more than the model. A few things I'd do from day one:
- Use a decent microphone: AirPods are fine, a desk mic is better, and the laptop mic in a noisy café is the worst case.
- Speak in full phrases: short fragments confuse punctuation. Say the whole thought, then pause.
- Turn on filler-word removal: the 1.7.1 setting strips the "um" and "uh" so you don't have to.
- Set your spelling variant: if you write in British English, choose it so "colour" doesn't get flagged on every pass.
- Build a custom dictionary: add names, product terms and acronyms early. The paid tier syncs these across devices, but even locally it saves time.
- Review before sending: no dictation tool is perfect. Skim the text once.
Dictate in thoughts, not in words. Say the sentence you'd write, pause, and let the app catch up.
A note on meetings
TypeWhisper's 1.6 release added calendar-aware meeting capture, and 1.8 builds are working on better meeting recordings. Recording a meeting is a different job from dictating an email, and there are consent rules to think about. In many places you need everyone's agreement before you record. Check before you hit the button.
Is TypeWhisper safe and private?
With local engines, your audio stays on your device, which is about as private as dictation gets. The project describes core dictation as designed to run locally without an account. The caveat is the optional cloud providers. The moment you route audio to Groq, OpenAI or Gemini, you're subject to their policies, not TypeWhisper's.
So the safe configuration is simple: local engine, no cloud provider, stable release. Because the code is open source, it's also auditable, which no closed competitor can offer. You still need to grant accessibility permissions on macOS, which is normal for any tool that types into other apps, but it's worth knowing you're doing it.
Should you pick TypeWhisper?
Yes, if you're comfortable with a little setup and want a free, private, flexible tool. Download the stable macOS build (1.7.1 at the time of writing), choose a mid-size local model, and give it a full week of real work before judging it. If it clicks, you've saved yourself a subscription. If the tinkering wears you down, that's a perfectly good reason to move to something more finished.
My verdict: on an Apple Silicon Mac, TypeWhisper is the best free dictation tool I know of, and the only one I'd trust with confidential text. Install 1.7.1 today, dictate your next three emails with a mid-size local model, and keep a note of every word it gets wrong. If that list stays short, you're done. If it grows, try GhostWriter next.
GhostWriter is a macOS app that turns your voice into clean, ready-to-paste text in any application, handling punctuation and formatting and adapting tone for things like emails or code editors. It's for Mac users who want the speed of dictation without choosing models or tuning settings. You can find it at justghostwriter.com.