How to Transcribe an Interview (Journalists & Researchers)
By the VidzTranscripts team · · 3 min read
Transcribing interviews by hand takes four to six hours for every hour of audio. Automatic transcription cuts that to a few minutes, but the result still needs care: journalists need quotes that are exactly right, and researchers need transcripts that are consistent, anonymised and ready for analysis. This guide covers the whole process, from the recording to the final checked transcript.
1. Get consent, and record it
Ask permission to record before you start and repeat the request once the recorder is running, so the consent itself is on the recording. Explain how the recording and the transcript will be used and stored. For academic research, follow the consent procedure approved by your ethics committee; many require written consent forms in addition to the spoken one. Laws on recording conversations differ between countries and states, so check the rules where both you and the interviewee are.
2. Record for transcription
- Use an external microphone if you can. A clip-on lapel microphone for each person gives far better results than one phone in the middle of a table.
- Choose a quiet place. Cafés are the classic mistake: the coffee machine and background chatter will cost you more editing time than the atmosphere is worth.
- For phone or video interviews, record each side on its own track if your tool allows it. Remote recording tools and most video-call software can do this.
- Make a backup recording on a second device for important interviews.
- Say the date, place and names at the start of the recording. It makes files easy to identify later.
3. Transcribe
Upload the recording and choose the interview's language. The transcript comes back with punctuation and a timestamp for every line. Download the version with timestamps: it makes checking quotes against the audio much faster.
4. Verbatim or clean?
There are two styles of transcript, and you should decide which one you need before editing.
- Clean (or intelligent) verbatim removes filler words such as "um", "you know" and false starts, and fixes obvious slips. It reads well and suits journalism, reports and most thematic research.
- Full verbatim keeps every word, pause, laugh and repetition. It is needed for conversation analysis, legal use and some linguistic research.
Automatic transcripts are close to clean verbatim: most filler words are left out and punctuation is added. If you need full verbatim, use the automatic transcript as a draft and add the hesitations and pauses while listening back.
5. Add speaker labels
Our transcripts don't label speakers automatically yet, but the paragraph breaks usually follow the changes of speaker, because people pause before replying. Go through the text once with the audio playing at 1.25× or 1.5× speed and add labels such as "Interviewer:" and "Participant 3:" at each change. For a one-hour interview this takes about fifteen minutes.
6. Check every quote
Speech recognition makes different mistakes from humans. It rarely mishears common words, but it can confidently write a similar-sounding word that changes the meaning: "can" and "can't" are a classic pair, as are numbers and names. Before you publish or cite a quote, find it by its timestamp and listen to it. This is non-negotiable for journalism.
7. Anonymise if needed
Research transcripts usually need names, places, employers and other identifying details replaced before they are shared or analysed. Replace them consistently, for example [Participant 3], [City] or [Employer], and keep the key that links codes to real names separately and securely. Remember that a combination of details (a job title, a town and an age) can identify someone even when the name is removed.
8. Analyse
Download the plain-text version to import into qualitative analysis software such as NVivo, ATLAS.ti or MAXQDA, or into a spreadsheet for simple coding. Keep the timestamped version alongside it, so you can always return to the audio for tone of voice and context that text alone can't carry.
How long it all takes
For a one-hour interview with clear audio, expect roughly 10–15 minutes for the automatic transcript, 15 minutes to add speaker labels, and 20–40 minutes to check and correct the text, depending on how exact it needs to be. That is around an hour in total, compared with most of a working day by hand.
Keeping recordings safe
Interview recordings often contain sensitive personal information. Store them in encrypted storage, share them only with people who need them, and delete them when your project's retention period ends. When you use VidzTranscripts, the recording is processed on our own servers, is never sent to third-party AI services, and is deleted as soon as the transcript is ready.