Interview
How to transcribe an interview: format, sample and an automatic way
To transcribe an interview, record it as an audio file, upload the file to get every phrase back with a timestamp, then add the speaker names by hand, because typing the whole hour yourself is commonly put at about four hours of work, while the automatic first draft is free for a first recording of up to 10 minutes. Below: the steps, verbatim against clean text, a sample, labeling and what it costs.
Updated
On this page
How to transcribe an interview in five steps
An interview goes from a recording to a readable transcript in five steps, and only the fourth one is manual.
- Record clearly and ask for consent. Put the microphone or phone close to the person answering, choose a quiet room, and say at the start that the conversation is being recorded; the rules on recording differ by country.
- Upload the recording. Drop the audio file on the upload area at the top. Without an account the first recording is transcribed free: in full up to 10 minutes, the first 10 minutes of a longer one.
- Check the text against the audio. Play the recording beside the transcript. A click on a timestamp moves the player to that line, so you only listen where the text looks doubtful.
- Add the speaker names. The automatic text has no speaker labels, so type “Interviewer” and the guest’s name in front of the lines wherever the voice changes.
- Export. Download TXT or Markdown for the finished interview, or SRT or VTT if the recording goes out as a video with subtitles.
A recording from a phone usually arrives as M4A, and the guide to transcribing a voice memo shows how to get it onto a computer. The upload accepts MP3, WAV and the other common audio formats too.
Interview transcript format: verbatim or clean
An interview transcript is written in one of two styles, verbatim or clean verbatim, and the choice decides how much of the speech survives in the text.
| Style | What it keeps | Fits |
|---|---|---|
| Verbatim | every word, including “um”, repeated words, false starts and notes such as [laughs] | research, legal and journalism work that must show how something was said |
| Clean verbatim | the meaning and the speaker’s words, without fillers and stutters | articles, case studies, quotes, podcast notes |
Speech recognition models tend to leave fillers out, so the automatic text lands nearer to clean verbatim. If your project needs every “um”, listen along with the player and add them at the timestamps that matter. If you want the opposite, a tidier text, an AI clean-up action works on the finished transcript.
An interview transcript sample
The automatic transcript of an interview is a list of time-stamped phrases, with nothing yet telling the questions from the answers.
Illustrative text in the exact TXT format, not a real recording:
[0:00] Thanks for making the time. Let's begin with how the project started.
[0:06] It started in 2019, in a garage, with two people and one borrowed laptop.
[0:14] And when did the first customers appear?
[0:18] About eight months later, mostly through word of mouth.After you add the names, the same lines read as a finished interview:
[0:00] Interviewer: Thanks for making the time. Let's begin with how the project started.
[0:06] Maria Lopez: It started in 2019, in a garage, with two people and one borrowed laptop.
[0:14] Interviewer: And when did the first customers appear?
[0:18] Maria Lopez: About eight months later, mostly through word of mouth.The stamp marks the start of each phrase, which is enough to find a quote again in the recording; the guide to how timestamping a transcription works compares it with other layouts.
Labeling speakers by hand
Labeling an interview means writing the speaker’s name in front of every line where the voice changes, and with timestamps in the text it is a quick pass.
- Use short, fixed names: “Interviewer” and the guest’s surname, the same way every time.
- Work with the player open: when a line is ambiguous, one click on its timestamp plays it.
- Keep the timestamps of quotes: they let a fact-checker find the exact second later.
- Use Markdown export for articles: bold names and one paragraph per turn convert easily.
Once the interview is labeled, the AI tools can build on it: an AI summary of the conversation for the key points, or an article draft from the transcript.
Time and cost of an interview transcript
An interview transcript costs one minute of balance per minute of audio, rounded up, and a first recording of up to 10 minutes is free.
- A 10-minute quick interview is your first recording: free, no account.
- A 45-minute interview uses 45 minutes, more than the 10-minute sign-up gift, so a pack covers the rest.
- A 60-minute interview is the 1-hour pack at $2.99.
The page shows the price before anything starts, and the balance is charged only after a successful transcription. The table of the pack prices lists all the packs.
What the automatic transcript does not do
The automatic interview transcript does not name speakers, does not separate overlapping voices and does not record the conversation for you.
- No speaker labels: you add them, as in the sample.
- Two people talking at once come out as a muddled line; check those spots in the player.
- Only the first audio track is used; two separate microphone files are uploaded one at a time.
- Files over 100 MB need sign-in, and files over 500 MB or 4 hours are turned away.
Frequently asked questions
How long does it take to transcribe an interview?
Typing one by hand is commonly estimated at about four hours per hour of audio. With an automatic first draft, that time goes into checking the text against the recording and adding speaker names instead.
Should an interview be transcribed word for word?
It depends on the use. Research and legal work need verbatim text with every filler and false start; an article or a blog post is better served by a clean version that keeps the meaning. Automatic recognition tends to skip fillers, so its draft is closer to the clean style.
Can the transcript label the interviewer and the guest automatically?
No. There are no speaker labels, only a timestamp on every line, so names are added by hand. The sample below shows the format.
Can I transcribe an interview for free?
Yes, the first recording: without an account it is transcribed free, in full up to 10 minutes and the first 10 minutes of a longer one. A whole hour-long interview after that uses minutes from your balance, and signing up adds 10 minutes as a gift.
Is my interview recording kept on your server?
No. The original is deleted as soon as the sound has been extracted from it, or after about an hour if the transcription never starts; only the text stays, in your browser or in your account. The privacy policy describes how files are handled.
What if the interview was recorded on a phone call or a video call?
Upload the recording as a file. A call saved as audio works as it is, and a video call saved as a video file works too, since the sound is taken from it.