Does editing a transcript change the original recording on iPhone?

Two different things share one recording

A voice memo is really two objects wearing one name: the audio itself, an AAC file written to disk while you were talking, and a transcript, a block of text produced afterward by asking a speech recognizer to listen to that finished file and guess at the words. They are created at different times, by different processes, and held as different data — the audio by the microphone and the encoder, the transcript by a separate pass that only starts once the recording already exists. Once transcription has run, the two sit next to each other in the same entry, but "next to" is doing real work in that sentence: neither one is embedded inside the other.

What editing a transcript actually touches

Correcting a transcript is a text edit in the plainest sense — the kind of change you would make in any editable field: select a word, replace it, move on. It does not call back into the audio at any point. Nothing decodes the .m4a, nothing re-encodes it, and nothing re-runs the speech recognizer against it. The correction updates one thing — the transcript text attached to that recording — and stops there. The audio finished being written the moment the recording ended; a transcript edit made an hour, a week or a year later has no path back into a file that closed long before the edit happened.

That is worth separating from a genuinely different action: re-transcribing a recording, which does call back into the audio to produce a fresh transcript from scratch. Re-transcribing reads the audio again. Correcting a word in the existing transcript does not read the audio at all — it changes text that is already sitting there.

Why the two are kept apart in the first place

A speech recognizer treats the audio it is handed as a source, not as something to modify — it listens once, writes out its best guess as a separate piece of data, and has no reason to ever touch the original recording again. Building it any other way would mean an app quietly editing its own audio on your behalf, based on a guess about what should have been there instead of what was actually recorded. Keeping the audio as a plain, unmodified file and the transcript as a separate, editable guess about that file is what makes it safe to fix as many words as you like: there is no way for a correction, or a hundred of them, to reach backward into the recording itself.

What "editable" does not mean: no version history

Editing a transcript overwrites the text in place rather than keeping a record of what it said before the edit. There is no history to open, no earlier draft sitting behind the current one, and no way to compare a correction against the recognizer's original guess after the fact. The same thing happens when a transcript is regenerated in a different language after picking the wrong one at the start — the fresh result replaces what was there rather than sitting alongside it. If the recognizer's original wording, mistakes included, is ever worth keeping for comparison, export the transcript before making changes you might want to undo; the export is a snapshot in a way the live transcript field is not.

Why this separation is actually useful

It means you can be as aggressive as you like about cleaning up a transcript without weighing that against any risk to the recording. Fix a name the recognizer guessed wrong, delete a line that got duplicated, tidy up a stretch of filler words — none of it is a decision that also touches the one copy of what was actually said. For anything where the original audio matters as a record in its own right — an interview you might need to check again later, a note you are transcribing loosely just to make it searchable — that gives you permission to edit the readable version freely while the audio sits underneath, untouched, as the actual account of what happened.

How this plays out with export

Because the transcript and the audio are separate, exporting one is a separate action from exporting the other, in every case, not just this one. Exporting the audio hands over the .m4a exactly as recorded, unaffected by any transcript edits made before or after. Exporting the transcript hands over whatever the text field currently says, corrections included. Doing one export tells you nothing about the state of the other. And if you have already exported a transcript once and go on to correct it afterward, whether that earlier copy catches up automatically is a related but separate question, worth reading about on its own.

How Voice Studio handles this

The transcript on a recording is ordinary editable text — correct a name, delete a line, tidy up filler words — and none of it touches the .m4a underneath, which is written once, when the recording finishes, and never rewritten by anything the transcript does afterward. There is no version history on transcript edits, so a correction replaces what was there rather than keeping it around for comparison; export the transcript first if the original wording is worth keeping. Recording, transcription, editing and export are unrestricted in the free app — the separation between text and audio is not a Pro feature, it is just how the two pieces of data have always worked.

Common questions

If I fix a mistake in the transcript, does it change the audio file?

No. Editing the transcript changes only the text attached to the recording. The .m4a audio file is written once, when the recording finishes, and nothing about editing the transcript afterward reads or rewrites it.

Can I get back the original transcript after editing it?

Not from inside the app — an edit overwrites the text in place, with no version history to revert to. Export the transcript first if the original wording, right or wrong, is worth keeping for comparison.

Is editing a transcript the same as re-transcribing a recording?

No. Editing changes existing text by hand and never reads the audio again. Re-transcribing runs the speech recognizer against the audio a second time and replaces the transcript with a fresh result.

Does exporting the transcript also export the audio, or the other way around?

Neither. They are two separate exports — the audio as the original .m4a, the transcript as TXT or JSON — and doing one does not touch or include the other.