Skip to content
How to · Mac voice writing

Voice to text without internet on a Mac: options that work offline

How to use voice to text on a Mac with no internet: Apple Dictation, local-model apps and command-line Whisper, plus what each needs online first and how to test.

Yes, you can use voice to text on a Mac without an internet connection, but you need to set it up while you are still online. There are three practical routes: Apple's built-in Dictation, which Apple says runs on the device for supported languages on Apple silicon Macs; a dictation app that runs a downloaded speech model locally; or command-line Whisper for transcribing recordings. Each needs a one-time download first. Then test with Wi-Fi off before you rely on it.

Publisher note: SuprFlow publishes this guide and is one of the products discussed. Competitor details come from each company's own pages, checked on 1 October 2026. Prices and features change; confirm before you buy.

What works offline, and what does each need online first?

OptionWorks offline forNeeds online firstCost
Apple DictationDictating into Mac apps, in languages Apple lists for on-device dictationTurning it on, and the speech-model download Apple's language page mentionsFree, built in
HandyLive dictation with a hotkeyDownloading a Whisper or Parakeet modelFree, open source (MIT)
VoiceInkLive dictationDownloading the app and a model; macOS 15.0 or laterSee its site; source is GPL v3
SuperwhisperLive dictation with a local voice modelDownloading a local model; other AI processing "needs the internet"See its pricing page
MacWhisper, direct downloadFile transcription with local models, and system-wide dictationDownloading a local model; cloud models are optionalSee its site
SuprFlowLive dictation with a local WhisperKit modelModel download, account sign-in and licence check$125 once, as of 1 October 2026
openai/whisper, whisper.cpp, WhisperKit CLITranscribing audio files from TerminalInstalling the tool and downloading a modelFree, open source

"Works offline" in this table means speech recognition runs on the Mac once the model is present. It does not mean the app makes no network connections at all; see on-device speech recognition explained.

Option 1: Can Apple Dictation work without internet?

On many Macs, yes. Apple's Mac User Guide for macOS 14 says: "On a Mac with Apple silicon, Dictation requests are processed on your device for supported languages—no internet connection is required." It also says that on an Intel-based Mac, or in a language without on-device support, "your dictated utterances are sent to Apple," and that in a search box, text "may be sent to the search provider."

To set it up (steps from Apple's macOS 27 guide):

  1. Choose Apple menu → System Settings, then click Keyboard in the sidebar.
  2. Go to Dictation and turn it on. Click Enable if asked.
  3. Read the text under the setting. Apple says it shows "whether your voice inputs and transcripts for general text Dictation ... are processed on your device and not sent to Siri servers."
  4. Check your language on Apple's feature availability page under "On-Device and Modeless Dictation," which notes "Requires download of speech models."

If your screen differs, search System Settings for "Dictation." Is Mac Dictation private? covers what Apple says about each case.

Option 2: Which Mac apps run a local speech model?

These apps run speech models on the Mac, per their own pages:

  • Handy describes itself as "A free, open source, and extensible speech-to-text application that works completely offline." It offers Whisper Small, Medium, Turbo and Large, and Parakeet V3, and runs on macOS, Windows and Linux. Models must be downloaded first; its README explains how to download them by hand on restricted networks.
  • VoiceInk's README says "100% offline processing ensures your data never leaves your device," lists macOS 15.0 or later, and uses whisper.cpp and Parakeet models.
  • Superwhisper's docs say "Superwhisper works offline. Local voice models transcribe without an internet connection," and that "AI formatting offline needs the local S1-mini model" in certain modes, while "Other AI processing needs the internet."
  • MacWhisper says "We use local models to transcribe your files," and also offers cloud models you can switch to. System-wide dictation is in the direct-download version; its support page says the App Store version does not offer it.

Whichever you pick, check that you have selected a local model, not a cloud one, and that any AI rewrite or formatting step is also local.

Option 3: How do you run Whisper offline from Terminal?

For recordings rather than live dictation, open-source Whisper tools run entirely on your Mac once installed.

  • whisper.cpp: the README downloads a model with sh ./models/download-ggml-model.sh base.en, then runs whisper-cli on a file. It notes whisper-cli "currently runs only with 16-bit WAV files."
  • openai/whisper: install with pip install -U openai-whisper and brew install ffmpeg, per the README. The package downloads each model file the first time you use that size, so run it once online with the size you plan to use.
  • WhisperKit CLI: brew install whisperkit-cli. Argmax's README says it "automatically downloads the recommended model for the device if not specified."

Run Whisper on a Mac has full steps. These tools print text in Terminal or write a file; they do not type into other apps.

How does SuprFlow behave without internet?

SuprFlow 1.0.0 runs speech recognition on the Mac with a WhisperKit (Whisper) model after you download it during setup. Its Default formatting is local cleanup of punctuation, capitalization and numbers, so it does not need a connection either. Cloud API formatting does need one, because it sends the transcript text to the provider you configure. Local (Ollama) formatting stays on the Mac only if the Ollama endpoint is on that Mac.

Speech running locally is not the same as the app never going online. Account and licence checks, app updates and usage telemetry are separate network traffic. Before relying on SuprFlow for a long stretch offline, read SuprFlow offline licence grace. Offline voice to text on Mac compares how the main apps handle local speech and formatting.

SuprFlow is not the right pick if you want it free (Apple Dictation and Handy are), need Intel Mac support (SuprFlow needs Apple Silicon and macOS 14 or later), or want to transcribe audio files offline (it does not; MacWhisper or whisper.cpp do).

How do you test offline dictation before you need it?

  1. Online: install or enable your option, choose the language, and download the model you will use.
  2. Quit and reopen the app, so you know the model loads from disk.
  3. Turn Wi-Fi off and unplug Ethernet.
  4. Dictate a fictional sentence into the app you really use: "Move the 3 March call with Lena Varga to 2:15 pm, and do not send the draft."
  5. Check the name, the time, the date and the "not."
  6. Restart the Mac with Wi-Fi still off and try again, if you will be offline for days.

If text appears with the network off, speech recognition is running locally. Note anything that stopped working, such as AI formatting or a sign-in prompt.

Questions people ask

Does Mac dictation work offline?

On an Apple silicon Mac, in a language Apple lists for on-device dictation, Apple says yes. On an Intel Mac, or in other languages, Apple says dictation is sent to Apple. Check System Settings → Keyboard → Dictation for your Mac.

What is the best offline speech to text app for Mac?

It depends on the job. For free live dictation, try Apple Dictation or Handy first. For transcribing recordings, MacWhisper or whisper.cpp. For a paid push-to-talk app with a personal dictionary, compare Superwhisper, VoiceInk and SuprFlow on your own voice. We have not run a controlled benchmark.

Do I need internet to download a speech model?

Yes, once. Apple's on-device dictation, every local-model app above and the command-line tools all download a model first. After that, recognition does not need a connection.

Will offline voice to text be less accurate?

Not necessarily. Accuracy depends on the model and your audio more than on where it runs.

Sources

Keep going

Read next

More how to
Put the guide to work

Make the first useful sentence.

Set up SuprFlow on your Mac, then try this guide in the app where you actually write.

  1. 01Download a speech model
  2. 02Dictate a harmless sample
  3. 03Check the text in its destination
SPECIAL OFFER$29 once Original price: $125

One license. Up to 5 Macs you own.Special offer, down from $125.Apple Silicon · macOS 14+

Get SuprFlow for Mac

7-day no-questions-asked refund on your first purchase. Contact support within seven days for a full refund. Read the refund terms ↗