Offline transcription on iPhone: how it works and what to look for (2026)

Yes, there are offline transcription apps for iPhone. Apple's Voice Memos and Notes transcribe their own recordings in ten languages on iPhone 12 or later, and some third-party apps run a speech model on the phone itself, so they turn audio into text in Airplane Mode without uploading it anywhere. This guide explains how that works and what to check before you trust one.

I make ScribeKit, one of those apps, so weigh what I say about it accordingly. The checklist below works for any app, and the Airplane Mode test at the end means you don't have to take my word for anything.

What is on-device speech to text on iPhone?

On-device transcription means the speech recognition model runs on your iPhone's own chip instead of on a company's server. The app records or imports audio, feeds it to a model stored on the phone, and gets text back. Nothing has to leave the device for that to happen.

A speech model is a large file of learned weights, usually a few hundred megabytes. On iPhone, apps typically convert that model to Apple's Core ML format so it can run on the Neural Engine and GPU, the parts of Apple's chips designed for exactly this kind of maths. The audio is turned into a spectrogram, the model predicts words and their timings, and the app stitches those into sentences or segments.

Two open model families dominate on iPhone right now:

• Whisper, OpenAI's open speech model, which covers about 100 languages and comes in sizes from tiny to large.
• Parakeet, NVIDIA's open model. The current version, Parakeet TDT 0.6B v3, is a 600-million-parameter model released under CC-BY-4.0 that covers 25 European languages and, in NVIDIA's words, "automatically detects the language of the audio."

A second distinction matters as much: where the model comes from. Some apps ship it inside the app, so they work from the first launch in Airplane Mode. Others download it after you install, which means one trip to the internet first. Both are offline once the model is on the phone.

If you want the engineering detail of one implementation, I wrote up how ScribeKit transcribes on your phone: 16 kHz audio in, timed segments out, no network in between.

Is there an offline transcription app for iPhone?

Yes. Transcription on iPhone falls into three groups, and only the first two are offline.

Apple's built-in transcription. Voice Memos and Notes show a transcript for recordings made in those apps, on supported iPhones and languages. It is free and already on your phone. Its limits are covered below.

Apps that run a model on the phone. These record or import audio and transcribe it on the iPhone's own chip, so they work in Airplane Mode once the model is on the phone. ScribeKit is one. Others exist, and the checklist further down is how to tell them apart.

Cloud apps that record offline. Many transcription services and "AI note taker" apps can capture audio without a connection, but the transcription happens on their servers, so you get text only once the audio has uploaded. They are not offline transcription apps, even if the recorder works on a plane. The Airplane Mode test tells you which kind you have in five minutes.

Why does offline transcription matter?

Privacy that doesn't depend on a policy. When a cloud service transcribes your audio, the recording sits on someone else's computer for at least as long as the job takes, and often much longer. What happens to it then is governed by a privacy policy you can't audit and that can change. The Freedom of the Press Foundation's review How secure are journalists' favorite transcription tools? goes through who can access uploaded files and whether recordings are used for training. Many people are fine with those terms. If you record therapy sessions, legal consultations, sources or anything else that isn't yours to share, an app that never sends the audio removes the question instead of answering it.

It works where the internet doesn't. On a plane, in a basement lecture hall, on a field trip, abroad without roaming. With a bundled model you can record and transcribe in Airplane Mode the same way you would at your desk.

No upload meters. Cloud plans tend to count minutes, because every minute costs the provider server time. On-device apps still have limits, but they come from the phone's memory and the maker's pricing, not from server bills.

No account needed to transcribe. An app that does the work on the phone has no reason to make you sign up, so there need be no profile for a transcript to be linked to. Whether a given app asks anyway is one of the things to check.

What are the trade-offs of on-device transcription?

Offline transcription is not free of costs. These are the honest ones.

App size. The model has to live somewhere. ScribeKit is about half a gigabyte because the Parakeet model ships inside it. Apps that download their model later look smaller in the store but take similar space once the model is in.

Battery and heat. Transcription keeps the Neural Engine and GPU busy for as long as it runs. For a long recording that is a noticeable, if temporary, drain, and the phone can get warm. Recording itself is light.

Older phones are slower. Speed depends on the chip, and some apps only support recent iPhones. ScribeKit runs on anything with iOS 18, but an older phone will take longer.

Speaker labels are rare and hard. Working out who said what (diarization) is a separate model and a hard problem on a phone. ScribeKit does not label speakers: you get timestamped segments, not "Speaker 1" and "Speaker 2". If speaker labels are essential, that narrows your choice a lot.

No meeting bots, rarely live captions. On-device apps are recorders, not services that join your video call. ScribeKit transcribes after you stop recording; it has no live transcript, no AI summaries, no Apple Watch app and no Mac app.

Accuracy depends on the model you get. A small model on an old phone will make more mistakes than a large one on a new phone. Language matters too: Parakeet is strong in its 25 European languages and does not cover the rest; Whisper covers more languages at the cost of speed and memory. My own comparison of the two lives in Parakeet vs Whisper on iPhone.

Does iPhone have built-in offline transcription?

Yes, within limits. Apple's Voice Memos guide (iOS 27 edition) says: "Audio transcription is available on iPhone 12 or later in English (all variants), Spanish, Portuguese, Italian, French, German, Japanese, Korean, Simplified Chinese, and Traditional Chinese. It's not available in all countries or regions."

In practice that means:

• iPhone 12 or later, running iOS 18 or later. An iPhone 11 runs iOS 27 but gets no transcript in Voice Memos.
• Ten languages, and no language picker. If you record in a language outside the list, or the phone detects the wrong one, you get no transcript or a garbled one.
• Only recordings made in the app. Apple documents no way to import an audio file into Voice Memos for transcription on iPhone. The Notes guide describes transcription for audio you record in Notes, "when the device language is set to" one of the same ten languages.
• Older recordings get transcribed when you open them: per Apple, recordings made in iOS 17 or earlier are transcribed automatically "if it includes recorded speech."

Apple's guide does not say whether the transcript is produced on the device, so I won't claim it either way here. If that matters to you, run the Airplane Mode test on Voice Memos yourself.

For English, Spanish or German on a recent iPhone, Voice Memos is a perfectly good free option. It runs out when you need another language, an older iPhone, or a file someone sent you. The step-by-step Voice Memos guide covers the missing-Transcript fixes, and the language list in detail covers the 19 European languages Apple doesn't transcribe. If what you want is to talk and have text appear in a message as you speak, that is dictation, a different feature; dictation vs transcription explains which one you need.

What to look for in an offline transcription app

App Store descriptions all say "private" and "fast." These are the questions that actually separate one offline transcription app from another. Ask them of any app, including mine.

Where does the model run, and is it in the app?

The transcription should happen on the iPhone, not on a server. Then check whether the model ships inside the app or downloads after you install. A bundled model works in Airplane Mode from the first launch; a downloaded one needs one trip to the internet first. Watch for apps that offer a menu of models where only some run on the device.

Which languages, and does it detect them?

Count the languages that work offline, not the total. Some apps cover more languages only with an extra download or a paid tier. Automatic language detection saves a step; a manual language picker helps with accents and speakers who switch languages.

Can it import files?

If your audio comes from another app, a recorder or a message, you need import from Files and the share sheet, and support for the formats you actually have (M4A, MP3, WAV). Check whether video files are accepted.

How long can a recording be?

Look for a per-recording limit and a monthly limit, on both the free and paid tiers. Larger models often have a lower cap because of the phone's memory.

Does it need an account?

An offline app has no reason to make you sign in to transcribe. If it asks for an account, find out what that account stores.

What does the privacy label say, and does it pass the Airplane Mode test?

The App Store privacy label is the developer's own answer about data collection. It does not tell you about network use. The proof is the Airplane Mode test: turn on Airplane Mode with Wi-Fi off, record a clip and import a file, and see whether text appears.

Does it label speakers?

If you need to know who said what, check whether the app labels speakers, and whether that also runs on the device.

How is it priced?

One-time purchase, subscription, free tier or a mix. Read what the free tier allows before you rely on it, and what is locked behind the paid tier.

How does ScribeKit answer those questions?

Here are my app's answers, as plainly as I can put them.

When the phone is online, ScribeKit makes an automatic request when it opens or returns to the foreground (rate-limited): it asks RevenueCat, its purchase service, whether you have Pro, sending an anonymous ID and purchase history, never audio or text. Downloading a Whisper model and buying or restoring Pro are the only other times it goes online, and you start those. You can record in the app (with pause and stop on the Lock Screen and in the Dynamic Island), and transcripts are split into timestamped segments you can tap to hear. It has no live transcript and no summaries.

Is there a free offline transcription app for iPhone?

Yes, with conditions:

• Voice Memos is free and already installed, for its own recordings, in ten languages, on iPhone 12 or later.
• ScribeKit is free for 10 transcriptions a month, up to 30 minutes each, in 25 languages, with the model built in. It also takes audio files and recordings from other apps (audio only, not video), and works on iPhones older than the 12.

If you have a computer and are comfortable with a terminal, OpenAI's open-source Whisper runs locally on it for free, with no limits beyond your computer's speed.

How can you check that a transcription app is really offline?

Don't take any app's word for it, including mine. Turn on Airplane Mode, make sure Wi-Fi is also off, record a short clip, import a file stored on the phone, and ask the app to transcribe both. An on-device app produces text; a cloud app shows an error, a spinner, or an upload queue. Apple's App Privacy Report then shows which domains each app contacted over the last seven days. The full protocol, with what a pass, a partial and a fail look like, is in Is your transcription app really offline?

Which offline transcription app should you use?

Pick by the one thing you can't do without:

• You only record in English or another of Apple's ten languages, on a recent iPhone, in Voice Memos: use Voice Memos. It's free and already there.
• You want it to work in Airplane Mode the moment it's installed, transcribe files from other apps, and start free: ScribeKit.
• You speak Polish, Dutch, Ukrainian, Swedish or another European language Apple doesn't cover: a Parakeet-based app handles all 25 Parakeet languages; Whisper covers more languages with heavier models.
• You need speaker labels: look for an app that says it labels speakers on the device, and test it. ScribeKit doesn't.
• You need live captions or a bot that joins meetings: that is what cloud services do, at the cost of uploading the audio.

Whatever you choose, run the Airplane Mode test before you trust it with a recording that matters.