How to transcribe an interview: a step-by-step guide

Transcribing an interview means turning the recording into a written text that you can read, code and quote. You either type it yourself while you listen or correct an automatic draft against the audio. Either way, you decide the level of detail, label the speakers, check every line and anonymise the transcript before you analyse it.

What does transcribing an interview involve?

The transcript becomes your data, so the choices you make while transcribing shape what you can analyse later. You decide how much detail to keep, how to mark speakers, pauses and passages you cannot hear, and how to remove names. Braun and Clarke’s Successful Qualitative Research (2013) gives transcription a chapter of its own, and their 2006 article on thematic analysis describes it as an excellent way to start getting to know your data.

Transcribing is also slow, careful work, and it is easy to underestimate. Plan it as part of your analysis, not as admin you do before the analysis starts, and agree the level of detail with your supervisor before you transcribe the first interview.

If you interviewed in another language, the usual approach is to transcribe in the language of the interview and translate only the extracts you quote, unless your supervisor asks for full translations. Say in your methodology who translated the extracts, and consider giving the original wording next to your translation so that readers can see what you worked from.

Manual or automatic transcription?

You can type every word yourself or let speech recognition write a first draft. A common middle way combines the two: an automatic draft, then a full manual check against the recording.

Braun and Clarke (2006) note that if someone else transcribed your data, you need to spend more time getting to know it, and that transcripts should always be checked against the original recordings. The same goes for an automatic draft.

How to transcribe an interview in seven steps

1. 1

   Prepare the audio

   Save a working copy with a code instead of a name, such as P07_interview.m4a, on storage your university approves. Listen to the first minutes to check the sound, and note the date, length, setting and speakers.
2. 2

   Choose a transcription style

   Decide between true and intelligent verbatim transcription or a more detailed notation, based on your method. Write a short key: how you mark pauses, laughter, overlaps and inaudible passages.
3. 3

   Set up a template

   Create one document layout for every interview: a header with the interview code, date, length, format and style, then one paragraph per speaker turn with a label such as INT or P07 and a timestamp. McLellan, MacQueen and Neidig (2003) set out practical considerations for preparing transcripts in this systematic way.
4. 4

   Transcribe

   Work in short chunks, slow the playback down and control it from the keyboard. If you use an automatic draft, rename the speakers first. Transcribe each interview soon after it takes place, while you still remember the context.
5. 5

   Check against the audio

   Listen to the whole recording while you read. Check names, technical terms, numbers, negations such as ‘can’ and ‘can’t’, and every change of speaker. Mark what you cannot hear as [inaudible] with a timestamp.
6. 6

   Anonymise

   Replace names, places, employers and other details that could identify someone with pseudonyms or bracketed descriptions such as [name of employer]. Keep the list of replacements separate from the transcripts.
7. 7

   Format for the appendix

   Add line numbers, keep the layout identical across transcripts, and put your transcription key at the start. Use only the anonymised version, name the file without real names, and check whether your programme wants every transcript in the appendix.

For anonymisation, the UK Data Service’s guidance on anonymising text data recommends replacing disclosive details with meaningful descriptors rather than deleting them, applying replacements consistently across transcripts and keeping an anonymisation log. Think about indirect identifiers too: a job title in a small organisation can identify someone without a name. A pseudonymised transcript is still personal data as long as anyone can link it back to a person, through the key or through details left in the text, and a recording is never anonymous, because a voice identifies the speaker.

How long does it take to transcribe an interview?

Rules of thumb circulate, but your own time depends on the sound quality, the number of speakers, accents and speed of speech, your typing speed and above all the level of detail. A true verbatim transcript takes longer than an intelligent verbatim one, and Jefferson notation many times longer again. Checking an automatic draft is usually quicker than typing from scratch, but it still means listening to every minute of the recording, often more than once.

The reliable way to plan is to time yourself. Transcribe ten minutes of one of your own recordings in your chosen style, including the check against the audio, and note how long it took. Multiply that by the total length of your interviews and add time for breaks, because few people can transcribe with full concentration for hours at a time. If you use an automatic draft, run the same test for correcting ten minutes of the draft.

Tools for transcribing interviews

For manual transcription, you need a player that slows the audio down and that you can control without leaving the keyboard. Express Scribe from NCH Software offers variable-speed playback, hotkeys and foot-pedal support, and has a free version for common audio formats. oTranscribe is a free web app in which, according to its website, the audio file and transcript never leave your computer.

• Your university’s tools: start here. Some universities license transcription tools or allow Microsoft Word’s Transcribe feature, which comes with some Microsoft 365 subscriptions, separates speakers and adds timestamps.
• Specialised services: online transcription services produce drafts with speaker labels and timestamps, and some offer human transcription for a fee. Check where each one processes and stores your data.
• Analysis software: some qualitative analysis programs offer transcription as an add-on, so the transcript goes straight into your project.

No automatic draft applies your transcription convention for you or catches every error. Spirelight, for example, gives you a draft with speaker labels and a timestamp on each speaker turn, in a ‘Cleaned up’ or a ‘Word for word’ style; for English recordings, Word for word keeps filler words and false starts. You still add pause marks, [inaudible] notes and laughter, anonymise the text and check every line by ear. The Word export has the speaker and the time range, such as [4:12 - 4:31], above each turn, and you can also export plain text or subtitles.

Check your university’s rules before you upload recordings anywhere: a voice identifies the speaker, so a recording is personal data even without names. Spirelight offers no data processing agreement; your university, not you, would sign one. Transcription runs on EU servers, files are stored with Google Firebase in the US, and AI features run on OpenAI in the US, which your ethics application and participant information sheet must state.

Where your university allows it, you can transcribe your interviews in Spirelight, rename the speakers, correct the draft while the audio plays and download it as a Word file for your appendix.

Invented example: a formatted transcript excerpt

The interview and excerpt below are invented. The study asks how international master’s students combine part-time work with their studies. The transcript is in intelligent verbatim, with timestamps at each turn and line numbers for citing.

In your analysis chapter you would cite the last answer as (P07, lines 6–8). The employer is replaced with a description that keeps what matters for the analysis, and the inaudible passage carries a timestamp so that you can listen to it again.

Consent, storage and recording

Your participants’ consent should cover the recording, the transcription, including any service that processes the audio, how long you keep the files and how you will use quotes. The guide to the participant information sheet explains what to tell them. Keep recordings and transcripts on storage your university approves, keep the pseudonym key apart from the transcripts, and delete the recordings when your approved retention period ends.

Many universities decide which apps and equipment are approved for interviews with personal data, and some approve neither phones’ built-in recorders nor third-party apps. Check your university’s rules before you record. Where your university allows browser recording, you can use the online voice recorder, which keeps the recording in your browser until you download it. Whatever you use, record a short test first and place the microphone close to both speakers.

After transcription: coding and analysis

With the transcripts checked and anonymised, read each one in full, ideally while you listen once more, and note first impressions and questions. Then start coding: label the passages that relate to your research question, either with codes developed from the data or with a framework you bring to it. The guide to inductive and deductive coding explains the difference, and reflexive thematic analysis shows how codes become themes.

Keep a note of every decision you made while transcribing, such as your style, your key and how you anonymised. You will need them for the methodology chapter, where readers expect to learn how the spoken interviews became the text you analysed.