
How to Transcribe an Interview for Research: A Step-by-Step Guide
Transcribe qualitative research interviews: consent, verbatim or clean style, participant codes, anonymization, accuracy checks and exports for coding.
To transcribe a research interview, first confirm that your consent forms and ethics approval cover how the audio will be processed. Then choose a transcription style that fits your analysis (verbatim or clean verbatim) and write it down as a convention. Transcribe with participant codes instead of names, anonymize identifying details, and check the whole transcript against the recording before you start coding. An automatic draft followed by a careful correction pass is the fastest reliable route.
This guide is for students and researchers doing qualitative interviews: theses, dissertations, user research and evaluation studies. Journalists can follow Interview Transcription for Journalists instead.
Before you record: consent and data protection
Transcription decisions start before the interview:
- Consent wording. If anyone other than you will process the audio, whether a transcription service, an assistant or a software tool, your participant information sheet should say so. Check the wording your ethics committee or IRB approved.
- Where the audio goes. Know which service processes the recording, where data is stored and how long it is kept. Your data management plan may require a particular region, an institutional tool or deletion after transcription.
- Recording quality. Use an external or clip-on microphone, record in a quiet room, and do a 10-second test. Clear audio saves more transcription time than any tool.
If you use FastScribe, uploaded audio on free accounts is kept for 12 hours and then removed. Pro accounts can choose 12 hours, 3, 7 or 30 days, or keep it until deleted. You can delete a job yourself at any time. See audio retention and the privacy policy before you decide whether that fits your protocol.
Step 1: Choose a transcription style that fits your analysis
| Analysis approach | Recommended style | What to keep |
|---|---|---|
| Conversation or discourse analysis | True verbatim with notation | Fillers, pauses, overlaps, intonation markers |
| Interpretative phenomenological analysis (IPA) | Verbatim or detailed clean verbatim | Hesitations and emotional cues that carry meaning |
| Thematic or content analysis | Clean verbatim | All words and meaning; drop "um", false starts |
| Applied or evaluation research | Clean verbatim | Clear statements you can quote |
Whatever you choose, write it in a short transcription conventions document. Include the style, how you mark pauses, laughter, inaudible words and overlapping speech, and how you anonymize. Supervisors and reviewers often ask for it, and it keeps transcripts consistent if more than one person transcribes.
Step 2: Set up participant codes and the transcript layout
Use codes from the start, not names:
- Interviewer (or I) for yourself.
- P01, P02 … for participants, numbered in interview order.
- A header with the code, date, interview length, interviewer and transcription style, but no real name.
A clean layout looks like this:
Participant: P07 | Date: 2026-11-12 | Length: 00:48:15
Interviewer: R. Chen | Style: clean verbatim | Transcribed: 2026-11-13
[00:03:12] Interviewer: How did the new rota change your week?
[00:03:17] P07: Honestly, it made nights easier, but handovers got longer.
We'd spend twenty minutes just finding out who had which patients.
[00:03:31] Interviewer: Can you give me an example?
[00:03:34] P07: Last Tuesday on [ward name], two of us thought the other had
done the medication round [laughs]. Nobody had.For more on layout, notation and timestamps, see How to Write a Transcript.
Step 3: Create a first draft
You can type the interview yourself or start from an automatic draft. Typing gives you deep familiarity with the data, but it takes many times the recording length. Most researchers now correct an automatic draft instead, then listen through once carefully, which still builds familiarity.
With FastScribe's interview transcription:
- Upload the recording. MP3, M4A, WAV and video files work, up to 1 GB.
- Choose the interview language. There are 105 options, including regional variants, which helps with accents.
- Check the speaker labels. Uploaded files are split by speaker. Rename them to Interviewer and P07 in the editor.
- Use the time range on each sentence to jump straight to any passage in the audio.
Without a subscription, a 45-minute interview uses 44 credits. New accounts receive 120 signup credits, and Pro includes unlimited transcription of uploaded files, which helps when a study has 20 or 30 interviews.
Step 4: Correct the draft against the audio
Listen to the full recording with the transcript open. Focus on the errors that change your analysis:
- Negation and hedging. "I'd never…" versus "I'd ever…", "kind of" versus a firm statement.
- Who said what, especially when the participant echoes your question.
- Domain terms, acronyms and names of services, drugs, places or software.
- Numbers and dates.
- Inaudible passages. Mark them as
[inaudible 00:14:02]instead of guessing.
If your methodology uses member checking, send the corrected transcript or a summary to the participant now.
Step 5: Anonymize
Work through the corrected transcript and replace anything that could identify a person:
| Original | Replace with |
|---|---|
| Participant's name | P07 |
| Other people's names | [colleague], [manager], [daughter] |
| Organization or ward | [hospital name], [ward name] |
| Specific place | [small town in the north] |
| Rare job title or unique event | A broader description that keeps the meaning |
Keep a key file that links codes to real identities separately from the transcripts, with access controls, or destroy it if your protocol says so. Anonymize both the DOCX and any other export, such as a spreadsheet.
Step 6: Export for analysis
- DOCX or TXT for import into NVivo, ATLAS.ti, MAXQDA or Dedoose. Most qualitative tools import Word and plain text files, and consistent speaker labels make speaker-based coding easier.
- XLSX if you code in a spreadsheet: one row per sentence, with time and speaker, ready to filter and add code columns.
- Keep the timestamps in your working copy so you can return to the audio when a passage is ambiguous.
FastScribe exports all three. See export formats.
Common mistakes in research transcription
- Changing style halfway through because time is running out. Decide the level of detail before the first interview.
- Cleaning up participants' language beyond your stated style. It can change meaning, and it misrepresents how people speak.
- Coding an uncorrected automatic draft. Small errors in negation or attribution lead to wrong themes.
- Storing names and transcripts together. Keep the key file apart from the data.
- Forgetting to note non-verbal cues that matter for interpretation, such as long pauses before a sensitive answer.
Summary
Good research transcription starts with consent and a written convention. Then: codes instead of names, an automatic draft corrected against the audio, careful anonymization, and an export your analysis software can read. FastScribe handles the draft, with speaker labels, a time range on every sentence, 105 languages and DOCX, TXT and XLSX exports, so your time goes into checking and analysis.
Frequently asked questions
Should research interviews be transcribed verbatim?
It depends on your analysis method. Conversation analysis, discourse analysis and some linguistic work need true verbatim with pauses and overlaps. Thematic analysis and most applied research use clean verbatim, which removes fillers and false starts but keeps the participant's words. Write your choice into your transcription conventions.
Can I use an automatic transcription tool for research interviews?
Often yes, if your ethics approval and consent forms allow audio to be processed by an external service. Check where the audio is stored, how long it is kept and who processes it, then correct the automatic draft against the recording before analysis.
How do I anonymize an interview transcript?
Replace each participant's name with a code such as P01, and replace other identifying details with bracketed descriptions, for example [hospital name] or [city in northern Italy]. Keep the key that links codes to real names in a separate, secured file, or destroy it if your protocol requires that.
How long does it take to transcribe a one-hour research interview?
By hand, several hours, depending on audio quality, the number of speakers and the level of detail. Starting from an automatic draft and correcting it is usually much faster, but plan time for a full listen-through before you start coding.
What format should research transcripts be in for NVivo, ATLAS.ti or MAXQDA?
Most qualitative analysis software imports Word documents and plain text. Use consistent speaker labels at the start of each turn, since some tools can detect them automatically. Check your software's import documentation for the exact formats it supports.
Categories
More Posts

How to Turn a YouTube Transcript Into a Blog Post That Actually Reads Well
A practical workflow for turning a YouTube transcript into a blog post with structure, examples, screenshots, and a clearer reader experience.

From 'Raw Audio' to 'Deep Insights': My Ultimate AI Workflow
Why your speech-to-text is only half the job. Discover how fastscribe.org + Perplexity AI creates the perfect content processing workflow.

UniScribe Review, Pricing, and Best Alternatives in 2026
Review UniScribe's current free limits, pricing, transcription features, and export formats, then compare practical alternatives for audio and video transcription.
Newsletter
Join the community
Subscribe to our newsletter for the latest news and updates