Offscript: Private Voice Memos

Offscript converts speech to text right on your iPhone or iPad. It downloads a language model once, then works without sending audio to a server. That's offline transcription in the strict sense: no upload, no server queue, no waiting on someone else's machine. You get transcripts that stay private, and the app runs even when you'd rather not rely on a connection.

Keyword search only helps if you remember the right word. What if you don't? Offscript lets you search notes by meaning, so a half-remembered idea still turns up. This works across the languages you transcribe in, which matters if you mix languages day to day. That's a real time saver.

You tag each note to keep a growing library tidy. Over time, tags become the fastest way to find a cluster of related memos instead of scrolling a long list, which gets painful once you've recorded hundreds of notes and can't recall what half of them were called. It's simple, and that's the point. No setup, no rules.

The app can lock everything behind Face ID, Touch ID, or your device passcode, which is a useful second layer on top of the on-device processing for anyone who leaves a phone on a shared desk or hands it to a friend. That extra step helps if someone else picks up your phone. Not ideal to skip if your notes are sensitive.

You can replay a recording while reading its transcript side by side. If a line looks off, you jump back to the audio and check. Handy for interviews and lectures where one word changes the meaning. Worth using every time.

Notes go out as plain text or Markdown. That keeps them usable in a notes app, a doc, or a publishing tool without conversion, which saves you the usual cleanup step that comes with exporting from voice apps. The export is clean enough to paste straight into your writing workflow.

Offscript transcribes in 99 languages with automatic detection, and the interface ships in 19 languages, so a note recorded in Spanish still turns up in a search you run in English later. If you speak more than one language, you don't have to switch settings for every recording. It just works.