Compare
Best private dictation apps for Mac, and how we checked
"Private" is not a regulated word. This page ranks apps by what they can be shown to do rather than what they say, and tells you how to repeat the check yourself.
Last updated
The four levels of "private"
Before any list, the categories — because vendors move between them without changing the word on the homepage.
| Level | What it means | How you verify it |
|---|---|---|
| 1. On-device | The audio never leaves the machine | Turn off the network and dictate |
| 2. Auditable on-device | As above, and the source can be read | Read the repository |
| 3. Zero retention | Audio is sent, transcribed and not stored | You cannot. You are trusting a policy and an auditor |
| 4. Encrypted and policy-bound | Audio is sent, stored, and governed by settings | You cannot. Read the policy and the defaults |
Levels 3 and 4 can be perfectly reasonable. They are not the same as levels 1 and 2, and only the first two can be checked by you rather than asserted at you. See what on-device actually means.
The apps, by level
Level 2 — on-device and auditable
VoiceInk. Says voice data never leaves your Mac
, with optional cloud enhancement handling transcribed text rather than audio — and publishes its source on GitHub so you can check. One-off from $25. Apple Silicon, macOS 14.4 or later. If auditability is your requirement, this is the top of the list and we are not going to pretend otherwise.
Level 1 — on-device
Vaitly Voice. Ours. Standard recognition runs on your Apple Silicon Mac and the audio is not uploaded. The internet is needed once, for the model download, and afterwards only for licence checks, updates, and an optional cloud rewrite that is off unless you configure it. A local history keeps your dictations, with audio, for a retention period you set.
Superwhisper. Says it works offline
, with the caveat that offline models run best on Apple Silicon. Also supports cloud models, so the default matters — check yours.
MacWhisper. Uses local models to transcribe your files
and processes sensitive content locally without data ever leaving your Mac
, with optional cloud AI connections.
Apple Dictation. On Apple silicon it can run on your Mac for supported languages. Apple tells you to check the Keyboard settings panel, because it varies. Free, and the right starting point.
Levels 3 and 4 — cloud, honestly described
Typeless states that audio is processed securely in real time and never stored in the cloud
, that dictation data is not used for training, and claims ISO 27001 certification with GDPR and HIPAA compliance. For a buyer whose requirement is a certification and a retention policy, that is a stronger position than anything on this page, ours included — we hold no certifications.
Wispr Flow states that transcription always happens in the cloud to provide the best speed and accuracy
, with data encrypted in transit … and at rest
and settings controlling storage and training. Clear and honest; simply a different category.
Verify it yourself in five minutes
- Complete first run with the internet on Local apps download a model once. Expected, and not the same as uploading audio.
- Turn Wi-Fi off properly Aeroplane mode, not just disconnected from a network.
- Dictate into Notes Text means local. An error or silence means it was going somewhere.
- Watch the connections Little Snitch, or
nettop -p <pid>. A local app is quiet while you dictate, apart from licence and update traffic. - Check the local history and the retention setting Local does not mean nothing is stored. Find where it lives and how long it is kept.
- Turn on FileVault It protects the local history whatever the app does. This is the single highest-value privacy action on this page and it is free.
If you have a compliance requirement
Be precise about which sentence you need to be able to say. "No audio leaves the endpoint" is satisfied by on-device recognition and can be demonstrated. "Processed only by a certified processor" is satisfied by a vendor with the certification and cannot be satisfied by us — we hold none, and our security page says so.
Those are different requirements and they select different products. See choosing dictation for confidential work.
Where these facts came from
Anything stated about another product was read on that product's own site or documentation on the date shown, and is re-checked at least quarterly. If something below has changed, tell us and we will correct it.
- VoiceInk says it "processes all voice transcription locally on your device" and that "your voice data never leaves your Mac," with optional cloud enhancement that handles transcribed text rather than audio. source, checked .
- VoiceInk is open source and published on GitHub for anyone to read, audit or build themselves. source, checked .
- Superwhisper says it "works offline, so you can transcribe anytime," and its FAQ adds that "Intel Macs work best with Cloud models. Offline models only run really well on Apple Silicon macs." source, checked .
- MacWhisper says it uses "local models to transcribe your files" and processes "sensitive content locally without data ever leaving your Mac," with optional connections to cloud AI providers. source, checked .
- Apple tells Mac users to 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," and notes this varies by language and region. source, checked .
- Wispr Flow states that "Transcription always happens in the cloud to provide the best speed and accuracy." source, checked .
- Typeless says "your audio is processed securely in real time and never stored in the cloud," and that dictation data "is never used for model training by us or any third party." source, checked .
- Typeless states ISO 27001 certification and GDPR and HIPAA compliance. source, checked .
Questions
Which Mac dictation app is the most private?
On verifiability, VoiceInk — it is on-device and open source, so the claim can be checked at the source. On architecture alone, any on-device app including VV, Superwhisper, MacWhisper and Apple's own. The most private setup is a local app plus FileVault plus a short history retention.
Does on-device mean nothing is stored?
No, and this is the most common misunderstanding. Most dictation apps keep a local history, often with the audio, so you can recover a dictation. That is on your disk rather than a server, which is better, but it is not nothing. Shorten retention on a shared machine and turn FileVault on.
Is a HIPAA-compliant cloud app better than an on-device one?
For a procurement process, often yes, because it produces a document. For the actual risk, on-device recognition means there is no transfer to govern. The two answer different questions, and which one you need depends on who is asking.
Can my dictation be subpoenaed?
Anything on a vendor's servers is subject to that vendor's jurisdiction and process. Anything on your own Mac is subject to yours. On-device recognition does not make data disappear — it changes who holds it, which is usually the point.