Home Guides How to transcribe an interview
How to transcribe an interview
Typing an interview out by hand takes roughly four to six hours per hour of audio. Automatic transcription brings that down to minutes, but what comes back is raw material, not a finished document: the work moves from typing to checking.
This guide covers the whole path — what to get right while you are still recording, how to get the text, and what to verify before you put a quote in print.
Transcribe for free60 minutes a month at no cost, no card and no sign-up to try it.
Step by step
- Record with the quotes in mindPut the microphone closer to the person you are interviewing than to yourself. A phone lying on the table between two people in a quiet room works; the same phone in a cafe does not.
- Upload the file or paste the linkMP3, WAV, M4A and OGG all work directly, and a phone voice memo is usually already M4A. If the interview was published on YouTube, paste the link instead.
- Read it once against the audioNames, job titles, company names, figures and technical terms are where automatic transcription fails first. Those are also exactly the parts you are most likely to quote.
- Cut it down to what you will useA full transcript is a search index, not a draft. Keep the passages you will actually quote and work from those.
What actually determines accuracy
Recording conditions matter far more than any setting. Two people at a similar distance from the microphone, in a room without music or traffic, transcribe close to perfectly. The same conversation in a restaurant loses whole phrases, and the person sitting further away loses the most.
If you only get one shot at the interview, spend the effort there: a quiet room and a microphone in the middle of the table are worth more than anything you can fix afterwards.
Speaker labels: you will be adding them yourself
The transcript comes back as continuous text. It does not mark who is speaking, so a two-person interview arrives as one uninterrupted flow of words. This matters more in an interview than anywhere else, because the whole point is attributing a sentence to a person.
A trick that costs nothing: at the start of the recording, say both names out loud, and say the interviewee's name again when you change topic. Those anchors make it much faster to split the text afterwards. Questions are also usually easy to spot, because they are the short turns.
From transcript to article
With an account you can turn the transcript into a summary or a draft web article directly, which is a faster starting point than the full text when the interview ran an hour.
One rule is worth keeping regardless: check every direct quote against the audio before publishing it. A summary is a useful skeleton, but the exact wording inside quotation marks is your responsibility, not the machine's.
Frequent questions
How long does a one-hour interview take to transcribe?
Usually a few minutes. The site reads the length of the file first and shows an estimate before it starts.
Can I get the transcript split by speaker?
No. The output is continuous text without speaker labels, so you separate the turns yourself.
Is it accurate enough to quote directly?
For clear audio, yes, but check the exact wording against the recording before publishing. Proper names and figures are the most common errors.
What if the interview is in another language?
It is transcribed in the language it was spoken in. It is not translated unless you ask for that afterwards.
Can I transcribe a phone or video call?
Yes, if you record it first and upload the file. Recording a call without telling the other person is illegal in many places, so ask first.