Since iOS 18, released in September 2024, Apple's Voice Memos app can transcribe a recording on the device without any third-party tool. That covers the simple case well. It does not cover the case most people actually want a few months later—finding the one recording out of two hundred where a particular thing was said—so this guide covers the built-in route first and then what it leaves unsolved.
Option 1: transcription inside Voice Memos
Open Voice Memos, select a recording, and look for the transcript control on the playback panel. On supported versions the text appears alongside the audio, scrolls with playback, and can be selected and copied like any other text. It runs on the device, so no upload is involved.
Two caveats before you rely on it. Apple has moved these controls between releases, so the exact tap sequence on your phone may differ from any screenshot you find online—look for the transcript or quotation-style icon near the waveform. And availability depends on both the iOS version and the language of the recording; transcription support is far broader for English than for most other languages.
Option 2: dictate instead of recording
If you want text and do not care about the audio, skip recording entirely. Tap the microphone key on the iOS keyboard in any text field and speak; the text arrives directly, and you can edit the wrong words while you still remember what you meant.
The trade-off is that you lose the original. For a shopping list that is fine. For anything where tone, hesitation, or an uncertain name matters, the recording is the evidence and the transcript is only an interpretation of it—and once dictation has guessed a name wrong, nothing remains to check it against.
Option 3: record into an app that keeps both
The third route is to record into an app that stores the audio and the transcript together and indexes the text across every recording you have ever made. That is the only one of the three that answers a question asked months later.
Concretely, that means a memo from March about a supplier problem is reachable in August from a description you invent at the time of searching, rather than from a filename you assigned in March or from remembering the exact words you spoke.
Why the built-in transcript is not an archive
Voice Memos organizes by recording. The transcript belongs to its file, the file has a name and a date, and everything you can do with it happens inside that one item. That is a fine design for a recorder and a poor one for memory.
- Search works on words you actually spoke, so a half-remembered topic finds nothing.
- Recordings sit apart from your typed notes and photos, splitting one week's thinking across three apps.
- Nothing extracts the commitment you made out loud, so ‘I will send it Thursday’ stays inside a sound file.
- Nothing connects a memo to the related note you typed two days earlier, because the app has no idea the note exists.
Getting a better transcript, whichever route you take
The largest gains are at recording time, not afterwards. Hold the phone within arm's length, avoid pockets and bags, and take three seconds to move away from the loudest thing in the room. Nothing downstream repairs audio the microphone did not capture cleanly.
Then help the recognizer where it is weakest. Say proper nouns slowly and consider spelling unusual surnames out loud; say numbers in full. Close with one sentence of context—who, what, and why this matters—because that sentence is what makes the memo findable when you have forgotten every other word in it.
Honest limits
On-device transcription is convenient and imperfect. Expect errors on names, technical terms, and any sentence that mixes two languages, and expect quality in your language to depend on how well that language is represented in the model. Numbers and dates are formatted by convention, so check any figure that carries consequences.
Nothing here is Apple-specific advice from Apple: iOS features move between versions, and the reliable instruction is what to look for rather than which pixel to tap. Android users have equivalent built-in options—Recorder on Pixel devices, Samsung's recorder on Galaxy phones—with the same underlying limitation, which is that a per-recording transcript is not a searchable history.
Use Voice Memos when you want the text of one recording, and dictation when you never want the audio. If the goal is to find something months from now, record into an archive that keeps audio and transcript together and searches every memo at once.