Offline dictation on Mac: the practical guide
How to dictate offline on Mac, step by step: enabling Apple Dictation's local mode, its real limitations, and when cloud or BYOK is the smarter call.
You're on a train, on a plane, or you're simply uneasy about your voice being sent to remote servers — you want to dictate without a connection. Good news: this is natively possible on Mac, and it only takes a few settings. Less good news: offline dictation has concrete limits that no app can magic away. This guide shows you how to enable it in two minutes, what it can do, and when a connected approach — or a smart compromise — makes more sense.
Yes, your Mac can dictate offline (locally)
Since macOS Sonoma, Apple Dictation downloads a speech recognition model directly onto your Mac for supported languages. Once that model is installed, you can cut the Wi-Fi: transcription keeps running, locally, on the Apple Silicon chip. No audio leaves the machine.
It's genuine offline dictation — free and built in. The only requirement is downloading the language module once (with a connection); after that, everything works without a network.
Enabling local dictation, step by step
- Open System Settings → Keyboard.
- Turn on Dictation, then accept the prompt.
- Under Dictation Language, choose or add your language (English, for example). macOS will then download the model.
- Wait a few minutes — the download happens in the background.
- Cut your connection to verify. Place the cursor in a text field, press the Fn key twice (or Ctrl depending on your setting), and speak.
If transcription works without a network, the local model is in place. Tip: a small microphone icon appears; as long as dictation responds without Wi-Fi, you're in offline mode.
The real limits of offline dictation
Let's be honest, because this is where most guides stop too soon. A model that fits on your Mac is necessarily lighter than a cloud model. In practice, expect:
- Poor punctuation. The local model transcribes words, but places commas, periods, and capitals badly. You often have to say 'comma' and 'period' out loud.
- No text cleanup. 'Ums,' repetitions, and hesitations appear as-is. Local dictation writes what you say, not what you meant to say.
- No context awareness. The same raw text gets inserted into Mail, Slack, or your code editor with no tone adjustment.
- Limited support for rare languages. Not all languages have a local model of equivalent quality.
The table below summarizes the three main approaches so you can see where offline stands relative to the rest.
| Approach | Connection required | Text quality | Privacy |
|---|---|---|---|
| 100% offline (local model) | No | Raw, weak punctuation | Maximum — nothing leaves the machine |
| Classic cloud | Yes | High, cleaned-up text | Low — audio goes to the vendor |
| BYOK (your own API keys) | Yes | High, cleaned-up text | Strong — under your own account |
Strict offline, or 'private but connected'?
Many people say 'offline' when what they really mean is 'without my data being sold or archived.' These are two different requirements.
If your need is absolute — nothing, ever, must leave the Mac, say for classified data or an air-gapped workstation — then 100% local is the only answer, and you accept raw text. Apple's local dictation, or an embedded Whisper model, does the job.
If your real need is control over your data without giving up clean text, there's a middle ground: BYOK (Bring Your Own Key). You supply your own API keys, audio goes directly to the provider you chose (OpenAI, Gemini, Groq), under your account and on your terms — no third-party vendor sitting in between or archiving anything. It's not offline — you need a connection — but it's far more private than an opaque cloud app, while recovering the punctuation and cleanup that local mode lacks.
To dig deeper into the privacy angle versus a popular cloud app, see our Speech Flow vs Wispr Flow comparison.
Where Speech Flow fits in
To be upfront: Speech Flow is not a 100% offline app. If your requirement is that nothing leaves the machine, stick with Apple's local dictation.
Speech Flow targets the other need: privacy without sacrificing text quality. It's a native macOS app (~50 MB, Apple Silicon). You hold Ctrl, you speak, and clean, punctuated text — with 'ums' stripped out — is inserted at your cursor in any app. An LLM rewrites on the fly and adapts the tone to wherever you're writing. In BYOK mode, you bring your own keys, audio goes straight to your provider, and no audio is stored. The app is multilingual (EN, FR, ES, IT). Full pricing details are on the pricing page.
FAQ
Is Mac offline dictation free?
Yes. Apple Dictation is built into macOS and local mode costs nothing. You just need to download the language model once with a connection.
Why is my offline text so poorly punctuated?
Because the local model is lightweight and focused on word recognition. Automatic punctuation and cleanup require a more powerful model running server-side — which means a connection.
Does Speech Flow work without an internet connection?
No. Speech Flow relies on remote models (your keys in BYOK mode, or the all-inclusive plan), so a connection is required. For strictly offline use, Apple's local dictation remains the right option.
In short: to dictate truly offline, enable Apple's local mode and accept raw text. If what you're really after is data control with a clean result, Speech Flow offers an honest BYOK path — native, lightweight, no audio retained. It's up to you to decide where to draw the line between absolute privacy and text quality; the details are on the pricing page.