How to redact a transcript before sharing it on iPhone

A voice memo is spoken the way people actually talk — a phone number read out because it was faster than typing it, a friend or a patient or a client named by name, a number or a diagnosis or a figure mentioned once in passing because the context made it obvious and harmless at the time. None of that is a mistake in the recording. It becomes a different question the moment the transcript of it is about to leave your phone as text someone can keep, search and forward on their own.

This is a different read from checking for errors

Reading a transcript to fix a mis-heard name is looking for what the recognizer got wrong. Reading it to redact is looking for what it got right and should not have written down for this particular recipient — a detail that is completely accurate and still does not belong in the message you are about to send. Those are separate questions, and a pass built for one tends to miss the other, because you are reading for different things each time.

What tends to show up that should not travel

• Someone else's name, mentioned in a context they never agreed to be part of — venting about a coworker, a story that names a patient or a client, a detail a friend told you in confidence.
• Numbers said out loud: an account number, a code, a password rattled off because dictating it was quicker than typing it.
• Anything medical, financial or legal that was fine to say in the room and is not automatically fine to hand someone as searchable text.
• A stray comment near the start or end of the recording — before you got to the point, or after you thought it had stopped — that was never meant for whoever receives the transcript.

That last one catches people more than the rest. A recording captures whatever is said while the microphone is open, not just the part you meant to send, and a transcript makes every word of it equally easy to read.

How editing actually works here

The transcript stays editable after a recording finishes — the same feature used to fix a mis-heard name works exactly as well for deleting a sentence outright. There is nothing specialised about redacting: select the text that should not go out and remove it, or swap in something neutral if the sentence needs to still make sense without it. It is ordinary text editing, done with the recipient in mind rather than the recognizer.

Because iOS speech recognition does not label who is speaking, a transcript of more than one person is continuous text with no "A:" or "B:" markers. A name mentioned twice will appear twice, in two different places, with nothing marking either one as special — so this is a read-the-whole-thing pass, not a find-and-stop-at-the-first-hit one. The recognizer also has no memory of what you removed the first time a name came up, so if it appears again later in the same recording, it needs the same treatment there too.

What redacting the text does not touch

Editing a transcript changes the text and nothing else. The recording underneath is untouched — the account number, the name, the detail you just deleted from the transcript is still fully audible in the .m4a file, exactly as it was said. If you send the audio alongside the redacted text, or instead of it, the thing you removed from the words is still there in the sound. Deciding what to redact only matters for the copy you are actually sending; if that copy is the audio, redacting the transcript alone does nothing for it.

The safest version of this, when the audio itself is not something the recipient needs to hear: send only the redacted text, and do not attach the recording. If the audio does need to go along with it — because tone matters, or because they need to hear the actual moment — redacting the transcript is only half the job, and the other half has no text-editing shortcut. The sound has to not be there in the first place, which usually means starting the recording after the sensitive part, or recording a separate clip that leaves it out entirely.

Your own copy versus what you send

Editing the saved transcript changes it everywhere from then on — in the app, in future exports, in the full backup. That is the right outcome if a detail genuinely should not be sitting in your own records either. It is not always what you want, though: if the sensitive line is only a problem for this one recipient and you would rather keep your own note intact, copy the parts you actually want to send and paste them into a new message instead of editing your saved transcript. Your original stays complete for your own reference, and only the excerpt you chose goes out.

This matters more for the full backup than it might seem. That export is one JSON file holding every recording's transcript exactly as it currently reads, meant for keeping everything, not for handing a redacted excerpt to one person. If the goal is a clean copy for someone else, exporting or copying that single recording's text is the right tool — the backup is doing a different job.

Nothing does this for you automatically

There is no feature, in Voice Studio or in iOS speech recognition generally, that flags a name, a number or a sensitive phrase for you. Speech recognition's job ends at turning sound into text; deciding what that text should not say to a particular person is a judgment call nobody automates, the same as reading over a document before forwarding it. Reading the transcript once, specifically for this, is the whole method — there is no shortcut that replaces actually reading it.

How Voice Studio handles this

The transcript is editable in the free app, with no limit and nothing held back for Pro — the same editing used to correct a mis-heard word works for removing a line outright. Export is TXT or the audio itself as the original .m4a, and either can be sent on its own or together, so the choice of what actually reaches someone is made at export time, not baked into the recording. There is no account and no server of ours in between, so an edit you make stays exactly where you put it — nothing is synced or uploaded anywhere beyond the export you choose to send.

Common questions

Does redacting a transcript change the audio recording?

No. Editing only changes the saved text. The .m4a audio file is untouched, so anything removed from the transcript is still fully audible if the recording itself is shared or played.

Can I redact a copy without changing the transcript I keep for myself?

Yes. Rather than editing the saved transcript, copy the parts you want to send and paste them into a new message. Your own saved version stays complete, and only the excerpt goes out.

Does Voice Studio detect sensitive information automatically?

No. There is no automatic detection of names, numbers or sensitive phrases, in this app or in speech recognition generally. Redacting means reading the transcript yourself and removing what should not go out.

If I redact a recording's transcript, does that change what is in the full backup export?

The full backup reflects whatever the saved transcript currently says, so an edit carries over. But the backup is meant to hold everything for your own records, not to serve as a redacted copy for someone else — export or copy that one recording's text instead when the goal is a clean excerpt.