Home · Personal · Tech & Digital Life · Files, photos & backups

YES

As of 13 August 2026, AI can transcribe an audio recording.

This still needs a person who signs their name to it.

Can you do it?

5 minutesto a draft.

30 minutesto something you’d act on.

Cost, all in£0

Skill needednone

Who has to check ityou

What the alternative costsThe listed commercial tools are AI photo, video or website products, so no comparable transcription alternative or price is supplied here.

If this goes wrong: words, names or speaker labels are wrong, and you need to replay the relevant part and correct the transcript before relying on it.

What to actually do

  1. Hand it to a person

    The route this page recommends

    A person who owns the outcome does this end to end, worth it when the failure is dear.

  2. Do it yourself

    Second choice

    A chat interface, no skill needed, and roughly 30 minutes until you can act on the result.

    How to actually do it

    1. Open a chatbot that accepts audio uploads and attach the original recording without editing or compressing it.
    2. Paste the transcription prompt and state the spoken language if it is not obvious from the file.
    3. Save the returned verbatim transcript and cleaned transcript as separate files so the original wording is not lost.
    4. Play the recording from the start and compare the transcript with it, correcting names, numbers, speaker labels and any passage marked [inaudible].
    5. Replay every unclear passage at a slower speed if your audio player allows it, and leave [inaudible] in place when you still cannot make out the words.
    6. Search the final transcript for names, places, organisations and technical terms, then compare each one with the recording or a source you already trust.
    7. Send or share the corrected transcript only after removing any audio or text that the other person is not entitled to receive.

    Prompt

    Transcribe the attached audio recording.

Return the result in this order:
1. A readable transcript with speaker labels where you can identify different speakers.
2. Timestamps for each speaker change or every clear topic change.
3. A list of passages that are unclear, marked [inaudible] rather than guessed.
4. A list of names, places, organisations and technical terms that may need checking.

Use the language spoken in the recording. Keep the wording faithful to the audio, including hesitations where they affect meaning, but remove meaningless repetition only in a separate cleaned version. Do not invent words, fill gaps from context or claim to identify a speaker unless the recording makes that clear. If more than one interpretation is possible, show the uncertainty. Provide both a verbatim transcript and a lightly cleaned transcript.

    Open it prefilled in ChatGPT or Claude, or copy it into Gemini, which takes no prefill link.

  3. Use a tool built for this

    The distant third

What it gets wrong

  • It cannot recover words that are absent from the recording or reliably resolve heavy background noise and overlapping speech.
  • It guesses at names, accents, numbers and specialist terms when the sound is unclear, even when the guess looks plausible.
  • It cannot know whether a speaker label is correct unless the voices or surrounding conversation make that clear.
  • It does not decide whether you have permission to upload someone else's voice recording or share the transcript.
  • It cannot replace a careful listen when the transcript will be used as an exact record.

Even on a YES, the friction has a name: verification cost and consent and privacy.

How we scored this

Five axes, each scored nought to two by hand: ten means AI carries the task cleanly, and the thresholds that turn a total into YES, PARTLY or NO are published in the methodology. Each axis name links to its definition.

AxisScore (0–2)
Output2
Inputs2
Verification1
Liability2
Effort delta2
Total9 / 10

The methodology and its thresholds are published in full.

FAQ

Can ChatGPT transcribe an audio recording?
Yes. Upload the recording and ask for a verbatim transcript with speaker labels, timestamps and [inaudible] markers instead of guesses. Listen back to unclear passages before relying on the result.
How accurate is AI transcription?
It is often usable for clear speech, but accuracy falls with background noise, accents, overlapping speakers, poor microphones and specialist vocabulary. Check names, numbers and every passage the model marks as uncertain.
Can AI tell who is speaking in a recording?
It can sometimes separate speakers and label them as Speaker 1 and Speaker 2. It cannot reliably identify people by name unless you provide that information or the recording makes their identities clear.
Is it safe to upload a private audio recording to AI?
Only upload it when you have permission and are comfortable with the service processing the file. Remove sensitive recordings where possible, check the service's current privacy settings, and do not assume that a transcript is private just because the recording is personal.

Nearby answers

All Files, photos & backups rulings

Assessed by gpt-5.6-luna (gpt-5.6-luna) on 2026-08-13, second-checked by an independent model. Wrong somewhere? Email [email protected] and it gets re-checked.

The newsletter

AI news, new answers and product picks, straight to your inbox.