TL;DR: A private dictation app is one where transcription happens on your own hardware, so your audio is never uploaded to anyone's server by default. Infina works exactly that way: NVIDIA's Parakeet model on the Apple Neural Engine, privacy mode on out of the box, $99 once as of July 2026, no subscription. And privacy does not cost you convenience: every other dictation app makes you press a key for each dictation, while Infina works hands-free from a couple of feet away. Say "type" plus your words and they get typed, say "send" and it presses Enter, say "open Notes" and you are dictating into the next app, no keyboard involved.

What "private" should actually mean in a dictation app

Marketing has stretched the word private until it covers everything from "we encrypt in transit" to "we promise not to look". Those are policies. A private dictation app should give you something stronger: architecture.

The test is one question. When you speak, does the audio leave your machine?

If yes, your words exist on someone else's infrastructure, under someone else's retention rules, subject to someone else's breach. If no, there is nothing out there to leak, subpoena, or train a model on.

Most cloud dictation vendors ask you to trust the policy. As one example, Wispr Flow's own data controls page says training on user dictation is on by default unless you enable Privacy Mode, as of July 4, 2026. The controls exist, but the default tells you whose interest the architecture serves.

Who actually needs a private dictation app

This is not paranoia. For a lot of professionals, cloud transcription is not a preference question, it is a compliance question.

  • Lawyers dictating case strategy, client names, and privileged notes. Privilege and third-party cloud processing make an uncomfortable pair.
  • Doctors and therapists speaking anything that touches patient information, where uploading audio to a consumer app can be a reportable event.
  • Founders and executives dictating term sheets, unreleased numbers, or acquisition notes that would move markets or leak strategy.
  • Anyone whose employer bans cloud transcription. Plenty of security teams block AI dictation tools entirely. An app where audio never crosses the network gives them nothing to block.

If you are in any of these groups, the default matters more than the settings page. You want privacy that holds even on the day you forget to check a box.

How Infina keeps dictation private

For a new user on an Apple Silicon Mac with the $99 license, here is exactly where your voice goes: nowhere.

  • Transcription runs fully on-device. Infina runs NVIDIA's Parakeet TDT 0.6B speech model on the Apple Neural Engine. Your audio never leaves your Mac. We wrote up the model itself in Parakeet dictation on Mac.
  • Text cleanup is on-device too, fast rule-based formatting, no cloud pass.
  • Hands-free listening is local. When hands-free mode waits for you to speak, that listening runs on-device, inside the app. Nothing is recorded or sent anywhere while it waits.
  • Privacy mode is on by default: no audio and no transcripts stored server-side. There is nothing to delete because nothing was kept.

Because nothing depends on a server, dictation also works with Wi-Fi off entirely, which we cover in offline dictation on Mac. Offline capability is privacy you can verify yourself: turn the network off and watch it keep typing.

Two precision notes, because absolutes are how privacy claims go wrong. If you manually turn privacy mode off to keep a dictation history, transcripts are saved to your account; that is your call, not a default. And if you opt into the cloud add-on, audio for those dictations does leave your Mac, which is the next section.

The cloud add-on: opt-in, labeled, and honest

Infina has an optional cloud add-on at $10/month, and we would rather explain it plainly than pretend it does not exist.

When you enable it, audio is sent, encrypted, to our cloud AI providers (Together AI and Groq) for sharper transcription, polished cleanup by large language models, and languages beyond English. That is cloud processing, full stop, and it only happens because you switched it on.

Cancel anytime and the app reverts to fully on-device. The default never changes underneath you: local first, cloud only by explicit choice, per dictation workflow you control.

This is also the answer to the polish question. Subscription apps charge $15/month forever with cloud processing as the only mode. Infina gives you their best trick, large-model transcription and LLM polish, for $10/month with its own 7-day free trial, on top of an app you own, and you can switch it off the moment a project turns sensitive. For the deeper architecture story, see on-device dictation on Mac.

Private and hands-free at the same time

Most privacy-first tools ask you to give something up. Infina's trade goes the other way: the private default comes with the one capability no other dictation app has.

Every dictation app on the market chains you to the keyboard, a hotkey pressed or held for every single dictation. Infina has push-to-talk too (hold Option), but double-tap Cmd and hands-free mode takes over.

From across the desk, coffee in hand, say "type draft a follow-up to the Meridian deposition summary". Infina types it. Say "send" and it presses Enter. Say "open Notes" and keep going in the next app.

Hands-free is off by default and labeled experimental, and the always-on listening that powers it runs on-device. To stay precise: some pieces of hands-free command handling are cloud-assisted, so the strictest zero-network guarantee belongs to push-to-talk dictation, the mode most sensitive work uses anyway.

No other dictation app completes that prompt, send, and switch-apps loop hands-free in plain English. Private by default and hands-free is a combination you currently cannot buy anywhere else.

The honest limits

  • English only in the base product. More languages require the cloud add-on, which means trading some locality for them.
  • Raw output by design. On-device cleanup is fast formatting, not a rewrite. Perfect for prompting AI tools, and when you want client-ready prose, the add-on polishes it.
  • Mac only, Apple Silicon required for the on-device models.
  • No free trial. Infina is $99 as of July 2026 with a 7-day no-questions-asked money-back guarantee, so you can test it against your own compliance bar risk-free. Details on the pricing page.

FAQ

Does Infina upload my voice recordings anywhere? Not by default. Transcription runs entirely on your Mac using the Parakeet model on the Apple Neural Engine, and privacy mode keeps audio and transcripts off our servers. Audio only leaves your Mac if you opt into the $10/month cloud add-on.

Is a private dictation app suitable for lawyers or doctors? On-device processing removes the third-party audio-upload problem that makes consumer cloud dictation hard to justify for privileged or patient-related material. Your compliance rules are your own, but with Infina's default there is no server-side audio or transcript to account for.

What happens to my transcripts after dictation? By default, nothing is stored server-side; the text just lands in the app you dictated into. If you manually turn privacy mode off to keep a history, transcripts are saved to your account, and you can turn that off again anytime.

Can my employer's firewall block Infina's dictation? Default dictation makes no network calls to transcribe, so there is nothing to block; it even works with Wi-Fi off. Only the opt-in cloud add-on and some hands-free command handling touch the network.

Is the cloud add-on a privacy problem? It is a labeled trade, not a leak. When enabled, audio goes encrypted to our cloud AI providers (Together AI and Groq) for sharper transcription and polish. It is strictly opt-in, cancel anytime, and the app reverts to fully on-device.

How much does Infina cost? $99 one-time as of July 2026, every 1.x update included, with a 7-day no-questions-asked money-back guarantee instead of a trial. The optional cloud add-on is $10/month with its own 7-day trial.

The bottom line

Ask one question of any dictation tool: does my audio leave my machine by default? "Yes, but you can adjust the settings" means your privacy is a policy. "No" means it is physics.

Infina is built on no. On-device transcription, nothing stored, cloud strictly as a labeled opt-in, plus the hands-free type, send, and switch-apps loop no other dictation app completes. $99 once as of July 2026, 7-day money-back guarantee, and your words stay exactly where you spoke them.