Quick answer: To transcribe interviews for qualitative research, listen to the recording and type out what was said, speaker by speaker, using a consistent set of conventions for pauses, laughter, interruptions and unclear words. Add timestamps at regular intervals, replace real names and identifying details with pseudonyms as you go, and check the finished transcript against the recording before you begin coding. A careful transcript is not just a record of the interview; it is the data you will actually analyse, so the time spent getting it right pays off at every later stage.
Transcription often looks like the mechanical part of a dissertation, but it shapes the analysis in quiet ways. A transcript that silently corrects a participant’s grammar removes evidence about how confidently they spoke, and one that drops every pause hides moments of uncertainty that might matter to your findings. This guide works through the choices step by step: choosing a style, setting conventions, planning your time, anonymising safely, and preparing the transcript so coding can begin smoothly.
Choose your transcription style before you type a word
The first decision is how faithful to the spoken word your transcript should be. There are three common styles, and the right one depends on your methodology and research question.
Full verbatim transcription captures every word exactly as spoken, including filler words such as “um” and “you know”, false starts, repetitions, pauses and laughter. It preserves the most information and is the expected standard for discourse analysis, conversation analysis and any study where how something is said is itself part of the evidence. Its drawback is readability: verbatim transcripts are long and harder to code quickly.
Intelligent verbatim transcription keeps every meaningful word but removes fillers, false starts and repetitions that add no meaning, and tidies obvious slips of the tongue. The participant’s meaning and voice remain, but the text reads cleanly. This is the most common choice for dissertations using thematic analysis, grounded theory or phenomenological approaches, where the content of what was said matters more than the mechanics of speech. If your supervisor has not specified a style, intelligent verbatim is usually the safest default.
Summarised transcription condenses the interview into its main points, paraphrasing rather than quoting. It is fast, but it is rarely acceptable as the primary record for a dissertation, because the researcher’s summary replaces the participant’s own words. Use summaries as memos alongside a transcript, not instead of one.
Whichever style you choose, record the decision in your methodology chapter and apply it consistently across all interviews. Switching styles halfway through a dataset makes transcripts incomparable and invites awkward questions at marking or viva.
Set your conventions and write them down
Write a one-page convention sheet before you start: rules for everything that is not a straightforward word. It prevents the slow drift of improvising a notation in interview three that you did not use in interview one.
Speaker labels are the foundation. Use a short, consistent label for each speaker, such as I: for the interviewer and P1:, P2: for participants, or pseudonyms once anonymisation is done. Start a new line every time the speaker changes, and never let two speakers share a paragraph. In a focus group, give every participant a distinct label and note in the transcript header who each label refers to, using pseudonyms rather than real names.
Timestamps anchor the transcript to the recording. Insert one at the start and at regular intervals, typically every two to five minutes, and at any moment you may want to revisit, such as a striking quotation. The format [00:12:34] is clear and easy to search. Timestamps also let a supervisor or second coder find a passage in seconds, and they are essential if you later need to check a quotation against the audio.
For the sounds and silences of speech, pick simple bracketed notations and use them every time. A short pause can be shown as (pause), a longer one as (long pause), laughter as (laughs), and an interruption with a dash at the point speech breaks off. Words you cannot make out should be marked (inaudible) with a timestamp, never guessed: a guessed word presented as fact is a data error, while an honestly marked gap is a limitation a reader can understand.
Non-verbal information deserves a line of its own in your conventions. Note visible emotion, such as a participant becoming tearful, and significant gestures where video or your field notes record them, in square brackets. Do not over-interpret; record what happened, not what you think it meant.
Prepare before you press record
Good transcription starts before the interview, with recording quality: a transcript can only be as good as the audio it comes from. Test your recorder in the actual room, check that it picks up a quiet voice from the far side of a table, and carry spare batteries or a charger. If you record on a phone, switch it to flight mode so a call does not interrupt the interview. Where possible, make a backup recording on a second device; a failed recording is the one transcription problem that cannot be fixed afterwards.
Consent covers transcription as well as recording. Your consent form and information sheet should state that the interview will be recorded, transcribed, anonymised and stored securely, and who will have access to the recording and transcript. If you plan to use a transcription service or automated software, say so, because sending a recording to a third party is a data-sharing decision your ethics approval needs to cover.
As soon as the interview ends, save the recording in two places, label the file with the participant code and date, and write your field notes while the conversation is still vivid. Note the setting, the participant’s apparent comfort, any interruptions, and topics that produced strong reactions.
Work through the recording in passes
Transcribing in passes produces far fewer errors than typing straight through in one sitting. Begin with a full listen without typing, or with only brief notes. This first pass gives you the shape of the interview, the participant’s main arguments, and the passages that will matter most.
On the second pass, type the transcript in short sections, replaying each stretch as often as needed. Work with headphones, use playback software that lets you slow the audio slightly and jump back a few seconds with a keystroke, and take breaks at least every hour: accuracy falls sharply when you are tired, and tired transcribers start tidying speech unconsciously. A realistic planning rule of thumb is to allow several hours of work for each hour of recorded interview, with verbatim transcripts taking longer than intelligent verbatim ones. Build that time into your dissertation timetable now, not the week before submission.
On the final pass, listen again while reading the transcript from start to finish, correcting errors and checking that speaker labels, timestamps and notations are consistent. This proof-listen catches the most common mistakes: a missing speaker change, a sentence attributed to the wrong person, and a quotation that reads smoothly but differs from what was actually said. Pay particular attention to numbers, names and technical terms, which are the easiest words to mishear and the most damaging to get wrong.
Automated transcription software can speed up the first draft. Treat its output as a rough draft, never as a finished transcript: automated tools regularly mishear accents, specialist vocabulary and overlapping speech, and they cannot apply your anonymisation or notation conventions. Always proof-listen the full recording against the automated draft, and check that your ethics approval and your university’s rules permit the tool you used, particularly where recordings are uploaded to an external server.
Anonymise as you transcribe, not afterwards
Anonymisation is easiest and safest when it happens during transcription, while the recording is in front of you and every identifying detail is visible. Leaving it to a later sweep means hunting through finished transcripts for names you may no longer recognise as identifying. Replace each participant’s real name with a pseudonym or code, and do the same for other people they name, workplaces, schools, hospitals, small towns and any other detail that could identify them in combination.
Indirect identifiers are the trap. A participant described as “the only female consultant in a rural practice” may be identifiable even with her name removed, because the description itself narrows the field to one person. Where a detail is identifying but irrelevant to your analysis, generalise it: “a small town in the north of England” rather than the town’s name. Where the detail matters analytically, discuss the risk with your supervisor.
Keep the key that links pseudonyms to real identities in a separate, password-protected file, stored apart from the transcripts and recordings, and never include it in an appendix. Delete recordings when your ethics approval and university policy require it, and state your storage and deletion plans in your methodology. Under UK data protection law, interview recordings are personal data, and participants’ rights over that data continue after the interview ends.
Get the transcript ready for coding
Before coding begins, give each transcript a header with the participant code, the interview date, the duration and the transcription style used. Number the lines or paragraphs if your analysis software or your supervisor’s feedback process uses line references; numbered lines make it possible to discuss an exact passage without ambiguity. Keep an untouched master copy alongside your working copy so that coding changes never overwrite the original record.
Read each finished transcript through once without coding, making brief familiarisation notes about striking ideas, recurring phrases and anything that surprises you. This reading is the bridge between transcription and analysis: it turns a typed record into data you have actually absorbed, and it often produces the first hunches that coding will test.
Once familiarisation is complete, the next step for most interview-based dissertations is thematic analysis: coding the transcripts systematically, grouping codes into patterns of meaning, and building themes that answer the research question. A clean, consistently formatted transcript makes that work far easier. Our guide to how to do thematic analysis walks through that next stage in detail.
From recordings to transcripts you can trust
If your interviews are piling up untranscribed, start with the earliest recording while your memory of it is strongest, and set a fixed daily transcription slot rather than waiting for a free weekend that never arrives. Work interview by interview, finishing the proof-listen and anonymisation of one transcript before opening the next.
United Academic Help supports students through the qualitative stages of a dissertation, from shaping interview questions and transcripts to coding and writing up findings. If you would like a second pair of eyes on your transcripts, help building a coding framework, or guidance on presenting quotations in your findings chapter, share your stage of work and your deadline and we will talk through the most useful next step for your project.
Frequently asked questions
How long does it take to transcribe an interview?
Allow several hours of work for each hour of recorded interview, with full verbatim transcripts taking the longest. Automated software can produce a rough draft more quickly, but proof-listening the whole recording against that draft still takes substantial time.
Should I transcribe interviews myself or use software?
Transcribing yourself gives you the deepest familiarity with your data, which strengthens later analysis, but it is slow. Software is a reasonable shortcut for a first draft if your ethics approval permits it. Either way, check the full transcript against the recording yourself.
What is the difference between verbatim and intelligent verbatim?
Verbatim transcription keeps every word, filler, false start and pause exactly as spoken. Intelligent verbatim keeps every meaningful word but removes fillers and repetitions that add no meaning, producing a cleaner text. Choose verbatim when how something was said is part of your evidence, and intelligent verbatim when the content is the focus.
How do I handle parts of the recording I cannot hear?
Mark them honestly as (inaudible) with a timestamp, and try replaying the passage with headphones or slowed playback before giving up. Never guess a word and present it as spoken. If a passage stays inaudible, note it in your methodology as a limitation.
Do I need to anonymise transcripts even for small studies?
Yes. Small studies carry the greatest identification risk, because a handful of details can single out a participant. Replace names, places and organisations with pseudonyms or general descriptions during transcription, and keep the linking key separately and securely.
When should I transcribe each interview?
As soon as possible after it takes place, and ideally before the next interview. Fresh field notes make notation decisions easier, early transcripts reveal whether your questions are working, and transcribing promptly prevents a backlog that collides with analysis deadlines.
Transcription rewards patience more than speed. Choose a style, write your conventions down, work in passes, anonymise as you go, and proof-listen every transcript against its recording. Do that for each interview, and the dataset you hand over to coding will be accurate, consistent and ready to support findings you can defend.