An offline transcription app that keeps your recordings on your iPhone.

What "offline" means here

Speech recognition in LocalBrain uses WhisperKit small, which runs on your iPhone. Text processing and source highlights use local language models that also run on the device. There is no cloud AI fallback and no cloud sync, and you do not need a LocalBrain account.

A few steps do use the network, and it helps to know which ones:

• Model downloads. The transcription model and local language models have to be downloaded before you can work offline. LocalBrain shows the download size and asks before it starts.
• App Store transactions. Buying or restoring Pro goes through the App Store and needs a connection.
• Web imports and search you choose. If you import a web page or run a web search yourself, that request goes online.

Everything else in the recording workflow works offline once the models are ready: transcribing audio, playing it back, editing and saving the transcript, generating source highlights, adding a source to your knowledge base and asking local AI questions about it.

Who needs offline transcription

Offline speech to text is useful whenever the connection is unreliable or the recording should not leave your phone.

• Meetings. Record a meeting or import one you already have, then check the transcript against the audio before you share anything from it.
• Interviews. Researchers and writers can keep raw interview audio on the device that recorded it, and quote passages with timestamps they can replay.
• Lectures and classes. Transcribe a lecture on the train or in a building with poor reception, then keep the parts worth studying.
• Sensitive recordings. Personal notes, private conversations or anything you would rather not upload can be transcribed without sending the audio to a transcription server.

How it works

1. Import. Record in LocalBrain, import a local audio or video file, or share an audio file from Voice Memos, Files or another app. Shared audio opens the Recordings / Transcript tab and is saved first. Video import extracts the audio track.
2. Transcribe. Start transcription yourself; receiving a file never starts it automatically. The first time, LocalBrain asks to download the required models.
3. Listen while editing. Play the recording and the current paragraph is highlighted. Scrolling, searching or editing pauses automatic following, and Return to playback position resumes it. Edit, then save. The original transcript is preserved.
4. Review source highlights. AI selects passages from the transcript as highlights, each with a timestamp you can tap to replay the audio.
5. Add to knowledge base. When a recording is worth keeping, choose Add to knowledge base. Nothing is added until you confirm, and the recording stays in its workspace.

For a full walkthrough starting from Voice Memos, read how to transcribe voice memos on iPhone. For the editing screen in detail, see how to edit an iPhone transcript while listening.

Offline vs cloud transcription

The main difference is where your audio goes and when you need a connection.

With cloud transcription, the audio file is uploaded to a server, processed there and the text is sent back. That needs a working connection for every recording, and a copy of the audio passes through the provider's systems under its own policies.

With on-device transcription in LocalBrain, the audio is processed on your iPhone. You need a connection once, to download the models. After that, each recording is transcribed locally, and the audio, transcript, edits and highlights stay in the app's local recordings workspace. Speed depends on your iPhone and the length of the recording, so a long file may take a while on the device.

Which approach fits depends on your situation. If you care most about keeping audio on the phone, or you often work without a reliable signal, local processing is the relevant option to look at.

Limits to know

• Free length limit. Free use transcribes up to 10 minutes per file. Active Pro removes that paid limit; device storage, memory and source limits still apply.
• System requirement. LocalBrain needs iOS 26.4 or later on a compatible iPhone with enough free storage for the model.
• First model download. You need a connection the first time to download the models. Do it before you head somewhere without signal.
• Recognition makes mistakes. Background noise, accents and overlapping voices affect results. Check names, numbers and dates against the recording before relying on them.
• No speaker labels. This release does not promise automatic speaker identification, calendar events, tasks or reminders.

If you mainly record meetings, listen, edit and save your meeting transcript shows the same workflow from that angle.

Common questions

Is LocalBrain a speech to text offline app after setup?

Yes. After the required transcription and local language models have been downloaded, transcription, transcript editing, source highlights, saved knowledge and local AI questions work without a connection. Model downloads, App Store transactions and web imports you choose still need network access.

Is my audio uploaded anywhere?

Audio, transcripts and notes are processed on your iPhone. LocalBrain has no cloud AI fallback and no cloud sync, and it does not require a LocalBrain account.

Which iPhones does it run on?

LocalBrain requires iOS 26.4 or later on a compatible iPhone, plus enough free storage for the selected model. The app shows the model download size before you agree to download it.

Can I transcribe a recording I already have?

Yes. Share an audio file from Voice Memos, Files or another app to LocalBrain, or import a local audio or video file in the app. The recording is saved first, and you start transcription when you are ready.