Private dictation apps for Mac: 7 privacy architectures compared

How this was put together

Method, scope, and a disclosure you should read first

Dictro is one of the seven apps compared here, and this page is published by Dictro. You should read it with that in mind. What keeps it useful anyway is the method: this is a documentation audit, so you can check every claim against the same public sources we did, including the ones about us.

One structural note that shapes the whole comparison: a privacy policy and a privacy architecture are different things. A policy is a promise that can be revised; an architecture decides what is even possible. An app that cannot reach a server cannot leak to one, whatever its policy says next year. That is the axis this audit sorts on.

The short version

Seven Mac dictation apps, compared on architecture

“Local by default” and “local only” are not the same guarantee. Several excellent apps here transcribe on your Mac until you switch on a cloud model — which is a real feature, but it means the privacy answer depends on your settings rather than on the app.

The evidence, app by app

What each vendor actually says

Dictro — on-device only, no cloud mode to leave on

Dictro runs speech recognition and its language cleanup pass on your Apple Silicon Mac. There is no cloud processing mode to opt into or forget to switch off, so the local guarantee does not depend on a setting. It works with Wi-Fi off, and finished text is placed straight into the field you were typing in.

What Dictro does send: in the interest of the same scrutiny applied to everyone else here — Dictro requires a sign-in, and the app sends an anonymous heartbeat containing a random identifier, the app and macOS version, and an aggregate count of words dictated. It never sends audio, transcripts, or the text you dictate. Transcript history is off by default and, if you turn it on, stays on your Mac. The full detail is in the privacy policy. Requires macOS 15 or later on Apple Silicon; English only today.

VoiceInk — local by default, and open source

VoiceInk states that “Your voice data never leaves your Mac,” and scopes its optional cloud feature carefully: “Cloud Enhancement is entirely optional — only transcribed text (not your voice) is processed if you choose to enable it.” It is also the only app here you can audit directly rather than trust: “VoiceInk is built in public on GitHub. Every line is out there for you to read, audit, or run yourself.” Source: tryvoiceink.com

superwhisper — local by design, with cloud models available

superwhisper's privacy policy is unusually direct: “Superwhisper is designed to process audio data locally on your device,” “all data is only stored locally on your machine,” and “Your data is not retained on Superwhisper servers.” It adds that it does “not collect any audio recordings, transcriptions, or personally identifiable information (PII).” Source: superwhisper.com/privacy

The nuance worth knowing: superwhisper also offers cloud AI models alongside offline ones, and its own FAQ notes that “Intel Macs work best with Cloud models. Offline models only run really well on Apple Silicon macs.” If you are on an Intel Mac, or you have selected a cloud model, check what you have configured.

MacWhisper — local by default, but a file tool, not live dictation

MacWhisper's documentation states that transcription “including Speaker Identification” is “all performed locally on your Mac or iOS device, and transcription data doesn't leave the local device.” It is also unusually clear about the exceptions: choosing a cloud transcription provider “sends the audio to a server owned by that cloud provider”; translation means “the transcription is sent to DeepL's servers”; and AI prompts mean “your transcription is sent for processing by an AI Service Provider,” unless you point it at a local runner such as Ollama or LM Studio. Source: docs.macwhisper.com

Note the category difference: MacWhisper is built around transcribing files and recordings, not typing into whatever app you are in.

Aiko — on-device, for recordings rather than dictation

Aiko “transcribes audio directly on your device, ensuring complete privacy,” running Whisper locally, and its own material suggests it “is perfect for sensitive recordings.” The scope limit is stated just as plainly: “The app does not yet support live transcription.” If your goal is text appearing at your cursor as you speak, Aiko is not that tool — but for private transcription of a recorded interview it is a strong one. Source: sindresorhus.com/aiko

Wispr Flow — cloud transcription with a genuine zero-retention mode

Wispr Flow is the clearest architectural contrast in this list, and it says so itself: “Transcription always happens in the cloud to provide the best speed and accuracy.” Its retention controls are real and worth crediting — with Privacy Mode on, “we keep nothing—no audio, no transcripts, no edits,” and none of your dictation data “is used by Wispr or any third party to evaluate, train, or improve AI models.” For a full opt-out it directs you to “Enable Privacy Mode and disable Private Cloud Sync for zero data retention.” Source: wisprflow.ai/privacy

The honest summary: Wispr Flow can be configured to retain nothing, but your audio is still transmitted and transcribed off your machine. Whether that matters depends on whether your concern is retention or transmission. For regulated or confidential material it is usually transmission.

Apple Dictation — free, built in, and configuration-dependent

Apple's guidance asks you to verify rather than assume. You can “check Keyboard settings to see whether your voice inputs and transcripts for general text Dictation…are processed on your device and not sent to Siri servers.” That on-device path is scoped to “general text Dictation (for example, composing messages and notes, but not dictating in a search box),” and Apple notes dictation “is not available in all languages or regions, and features may vary.” Source: Apple Support

It is free and already installed, which makes it the right baseline to try first. Its limits are punctuation, filler words and formatting — it transcribes fairly literally rather than cleaning up speech.

Matching the tool to the threat model

Which one to pick

• You want live dictation and no cloud mode to worry about: Dictro. There is no setting that sends audio away, because there is no cloud path.
• You want to verify the code, not the promise: VoiceInk — it is the open-source option here.
• You want the most feature-rich local dictation: superwhisper, with your model selection checked.
• You are transcribing recordings and interviews: MacWhisper or Aiko — file tools, not cursor dictation.
• You want maximum accuracy and accept cloud transmission: Wispr Flow, with Privacy Mode on and Private Cloud Sync off.
• You want free, and literal is fine: Apple Dictation — check the on-device setting first.

Trust, then test

The two-minute test that settles it for any app

Every claim on this page is a quote from a vendor. Here is the one check that needs no vendor at all, and it works on any dictation app you are evaluating.

1. 1

   Turn off the network

   Switch Wi-Fi off from Control Center and unplug any wired connection.

2. 2

   Dictate into a local app

   Put the cursor in Notes and dictate a sentence with the app you are testing.

3. 3

   Read the result

   Text with no connection means the transcription ran on your Mac. An error or an empty field means it needed a server.

4. 4

   Repeat with your real settings

   For apps with optional cloud models, test the configuration you actually use — not the default you installed.

For Dictro's own architecture in more depth, see private dictation for Mac, the offline dictation guide, or the Mac dictation app overview.