Best private dictation apps for Mac

Pros

• The vendor says no AI models are trained with user data.
• Optional local history is stored AES-256 encrypted.
• Self-described as GDPR compliant; the operator is Akilaris UG in Bremen, Germany.
• Licensing and website servers are in the EU, per its homepage.

Cons

• No third-party certifications are stated.
• The homepage and download page disagree on Apple silicon only versus an Intel build, and the homepage and terms disagree on currency.
• The AI Cleanup model is described only as 'a local model' and is not named.

Pros

• Source code is public on GitHub under GPL v3, so you can read it or build it yourself.
• Local models run after download, and its privacy policy says transcripts and audio are not stored on VoiceInk servers.
• Screen context is read with local OCR, and the screenshot itself is not sent.
• Ollama can run text cleanup locally.

Cons

• AI enhancement sends text, plus any context you enable, to the provider you pick.
• Local is set per Mode, so each Mode's model and provider needs checking.
• Activating a purchased licence needs a network connection.
• Minimum macOS is 14.4 on its website but 15.0 in the GitHub README.

Pros

• The core app is open source under GPLv3.
• Dictation works offline with downloaded local speech models.
• API keys for optional providers are stored in the macOS Keychain.
• An optional Fluid-1 cleanup model runs on the Mac.

Cons

• Fluid Intelligence, the optional local cleanup runtime, is private and not open source.
• Anonymous usage analytics are on by default, with an opt-out. Voice, audio and transcribed text are stated as not collected.
• Needs macOS 15.0 or later, and Apple silicon for most models.
• No certifications or compliance attestations are stated.

Pros

• Works offline with the free local speech models.
• Its backend does not store audio, per its privacy policy.
• The Local Models plan is $0 forever and needs no account.

Cons

• AI Instructions can send the transcript and optional context to AI services, even when transcription ran locally.
• Third-party transcription providers may keep data for their own purposes, per its privacy policy.
• Operated by LLP Inviral Tech, a company registered in Kazakhstan, per its terms.
• No security certification is stated on the pages we read.

Pros

• Local voice models work offline. Offline AI formatting needs the local S1-mini model in Message or Mail mode.
• States that audio and text sent through cloud models are not used to train AI models, and that history stays on the device.
• Enterprise customers can get a SOC 2 report under NDA, plus a signable DPA and BAA.
• The Free plan keeps unlimited local Whisper dictation.

Cons

• Each mode chooses its own models, so staying local depends on your mode settings.
• AI processing other than local S1-mini formatting needs the internet.
• Its Privacy Policy page is dated June 2024, before the cloud models, so rely on its newer docs for current processing.

Pros

• Built into macOS, so there is nothing to install.
• Apple says audio is not stored unless you opt in to Improve Siri and Dictation.
• On-device processing covers a published list of languages: 48 language and region entries on the macOS 27 page, as we counted them.

Cons

• Whether you get on-device processing depends on your language and settings, so check Keyboard settings. Otherwise audio goes to Apple servers.
• Apple's privacy page says it may retain request history, including transcripts, for up to two years to improve Siri, Dictation and related models.
• Apple's macOS 15, 26 and 27 pages do not say Dictation works without internet and say you may need a connection. Its Ventura 13 and Sonoma 14 pages say no internet is required on Apple silicon.