A private dictation app with no account

A private dictation app transcribes and cleans up your speech on your own device instead of a server, with no account required to use it. Parlo works this way on iPhone and Mac, with on-device transcription, encrypted history, and end-to-end encrypted sync you can turn off. The beta is free.

By Eric Friedman. Last updated 2026-09-08.

Type "private dictation app" into a search bar and you get a mix of things. There is the built-in dictation on your phone, a handful of paid apps that promise privacy, and a couple of on-device engines you install yourself. Most of them use the word "private" a little differently. Some mean your recording never leaves your device. Some mean it leaves your device but nobody stores it afterward. Some mean you can turn on a local mode if you dig into settings, while the default is a server somewhere.

None of that is dishonest by itself. It just means "private" is a claim to check, not a label to trust. This guide walks through what the claim can mean and what to look for on any vendor's own site. It also covers how the main apps in this space answer those questions today, and where Parlo, the app behind this site, fits into the picture.

Six players are worth knowing before you compare anything. Apple's own dictation is built into iPhone, iPad, and Mac at no separate cost. Wispr Flow is a cloud transcription app across four platforms. superwhisper and Spokenly both let you choose between an on-device model and a cloud one. Monologue by Every is a Mac and iPhone app that defaults to the cloud with an on-device option. Parlo runs on-device only, with no cloud step to opt out of.

This guide is for you if you dictate anything you would rather not hand to a server. That could be client notes, a draft of something unfinished, or a message you are still deciding how to word. It is also useful if you simply want to know what a privacy claim on a product page is actually promising before you install anything. None of the checks below require you to trust a marketing page. Each one points at a specific sentence a vendor either states or leaves out.

The four checks for a private dictation app

Before you take any vendor's privacy page at its word, there are four questions worth answering yourself, using their own documentation.

Where does the audio actually go? Some apps transcribe entirely on the device. Others send the recording to a server and delete it afterward. Others let you pick, with cloud as the default and a toggle for local mode. The difference is not academic: once audio leaves your device, its handling depends on that company's servers, retention policy, and any subprocessor it uses.

Does the app need an account? An account is not automatically bad, but it usually means the vendor can tie your dictation activity to an identity it controls. A dictation app that works the moment you install it, with no sign up screen, has fewer places for that link to form.

What happens to your history? A transcript is often more sensitive than the single sentence you just said. It is a running record of what you have dictated over months. Ask whether history is stored at all, where, and whether it is encrypted in a way the vendor itself cannot read.

What does a claim like "processes locally" or "works offline" actually cover? Read past the headline. Local processing for transcription does not always mean the cleanup step is local too. An offline claim does not always mean the first run skips a one-time download of a language model. A private dictation app should be specific about which steps are on-device and which, if any, are not.

Wispr Flow answers the first question directly on its own privacy page. Its transcription runs in the cloud to keep speed and accuracy high. Cloud storage of your transcripts and audio is a setting you turn on rather than something that ships on by default. superwhisper and Spokenly both let you choose, with a local model as one option and a cloud model, using your own key in superwhisper's case, as the other. Monologue's default path sends dictation to its AI providers under a stated no-retention policy, with a separate on-device option for Mac. Apple's own dictation can run on-device or be sent to Apple's servers, and the device tells you in Keyboard Settings which one is active for you right now.

How the main apps compare

The table below uses only what each vendor states on its own site, checked on 2026-09-08. Where a page does not say, the cell reads "not stated" rather than a guess.

If a local Wispr Flow alternative is specifically what you are after, the dedicated comparison goes deeper on price and platform differences between the two.

Where does my audio go?

This is the question worth asking before any other. A dictation app has to capture audio to do its job, so the real question is what happens to the recording after that. On an app that processes on-device, the audio is converted to text locally and there is nothing left to send anywhere. On an app that sends audio to a server, the vendor's privacy policy is the only place to learn what happens next. That policy is where you find how long the audio is kept, whether it trains a model, and who else can see it.

Watch for absolute claims here. A vendor that says its app works offline is telling you about the transcription step, not necessarily about setup. Most on-device speech engines, Apple's included, need a one-time download of a language model the first time you use them, which needs a connection once. After that, the everyday work of turning speech into text does not need one. A private dictation app should be honest about that first-run step instead of promising an experience with no exceptions. What on-device recognition means, engine by engine, is covered in on-device speech recognition explained.

It is also worth asking what happens to the finished transcript, not only the audio. A vendor can process speech entirely on-device and still upload the resulting text for a cloud-based cleanup or a synced history feature. Look for both halves of the pipeline, transcription and whatever runs after it, before deciding an app is private end to end.

What does Full Access actually allow?

An iPhone dictation app that ships as a keyboard, rather than a separate app you copy from, will ask for Full Access the first time you enable it. Apple's own documentation for custom keyboards explains what the setting technically permits. A keyboard extension with Full Access can share data with its companion app and, depending on what the developer builds, reach the network on its own.

What matters is what a specific app actually does with that access. A keyboard extension cannot record audio under any permission level, since iOS reserves the microphone for the containing app. A private dictation app built as a keyboard has to record and transcribe in its main app instead. It then hands the finished text to the keyboard through a shared, sandboxed container. Check whether that keyboard also talks to a sync service on its own. A keyboard that stays off the network entirely is a stronger privacy position than one that can reach it.

What does a private dictation app cost?

Pricing in this category ranges from free with limits to a monthly subscription. Wispr Flow's free tier is capped by word count and its Pro plan runs $15 a month, or $12 a month billed yearly. superwhisper charges $8.49 a month, $84.99 a year, or a one-time $249.99 for a lifetime license, after a free tier with limited access to its cloud models. Spokenly's local models are free with no usage limit, while its Pro tier, which covers Spokenly's own cloud models, runs about $8.33 a month or $99.99 a year. Monologue's free trial gives a capped word and note allowance before its $15 a month or $144 a year Pro plan takes over.

None of that is unreasonable for a subscription product. It is simply a different model from a free beta with no account and no purchase code anywhere in the app. Weigh that difference against the processing question above, not instead of it.

Is a beta safe to rely on?

A beta means the software is still being finished, not that it is unsafe to use every day. Expect rough edges and a faster pace of change than a shipped product, and check that your data stays protected even if a particular build has a bug. History that is encrypted at rest on the device, and encrypted again before it syncs, stays protected regardless of what a bad release does elsewhere in the app. Send anything that goes wrong to the feedback address so it gets fixed for the next build.

Testers use dictation, cleanup, and the Mac hotkey as part of a normal workday, not only to try the app once. Treat a beta label as a reason to report problems, not a reason to keep your real work out of it, once the privacy questions above are answered.

Where Parlo fits

Transcription uses Apple's SpeechAnalyzer and SpeechTranscriber, running on your iPhone or Mac. It is the only engine Parlo uses. There is no cloud transcription step to opt into or out of, because there is not one at all.

Audio itself is never kept. Parlo captures it, turns it into text, and discards it. Nothing about a dictation persists as a recording, on the device or anywhere else.

Your local history is encrypted at rest, under a key that lives only on the device and never rides along in a backup or iCloud Keychain export. Losing the device without that key means the local history cannot be recovered either, which is the tradeoff of a key nobody but you holds.

History sync is on by default, and the encryption happens before anything leaves the device, using a key kept in your iCloud Keychain. Apple and Parlo hold only ciphertext, never the plaintext or the key, so a synced dictation cannot be read by either one. Turn sync off in Settings at any time, or skip signing in with Apple and nothing syncs to begin with. How the end-to-end encrypted iCloud sync works, key and all, is its own explainer.

There is no Parlo account to create in the first place. Signing in with Apple is an optional identity for sync, not a login required to dictate. Signed out, every part of the app works, and your dictations stay local only. The voice to text without an account article walks through what an account-free app does and does not give up.

On iPhone, the Parlo keyboard cannot record audio, since iOS reserves the microphone for the app itself. Dictation happens in the Parlo app, which hands the finished text back to the keyboard through the shared App Group. The keyboard never links the sync code and never talks to iCloud on its own.

The first run of the app may need a one-time download of Apple's on-device speech model, which needs a connection once. After that, dictation and cleanup run without one, aside from the beta's daily prompt refresh, which carries no dictation content.

The beta itself costs nothing. Both apps contain no purchase or subscription code at all. Install Parlo for iPhone through the public TestFlight link, or Parlo for Mac from the notarized download on the beta page. There is nothing to buy and nothing to sign up for.