Interview transcription is the process of converting a recorded conversation between a researcher and a participant into a written transcript. In qualitative and market research, an accurate transcript preserves the participant’s words so researchers can review evidence, compare interviews, identify themes and return to the original audio when interpretations need checking.
A research-ready transcript is more than a document with correct spelling. Its style, speaker labels, timestamps, uncertain passages and treatment of identifying information must match the study’s methodology and ethical requirements. This interview transcription guide shows how to prepare, transcribe, check and organise interview recordings, with an example you can adapt for an in-depth interview (IDI).
For a finished transcript rather than DIY instructions, see interview transcription services. If your recordings are confidential, discuss the project specifications and approved transfer arrangements before sending source files.
What is research interview transcription?
Interview transcription records what is spoken during a one-to-one interview, semi-structured interview, oral-history interview, market research IDI or similar research conversation. The transcript can help a team locate particular statements and examine language more systematically than repeatedly listening to a long recording.
Researchers should distinguish three activities: transcription reproduces the recorded conversation according to an agreed convention; quality assurance verifies the transcript against the source and flags uncertainty; and coding or thematic analysis interprets the data. Coding can use the transcript as an input, but is not itself transcription, and the coding choices should remain traceable to the source.
Interviews differ from focus groups, where several participants may overlap and speaker attribution can be especially challenging. This page concentrates on interviews; a separate qualitative interview coding guide covers what happens after transcription.
Choose a transcription style before recording begins
Transcript conventions should follow the research question and analysis plan. A discourse-analysis study may need speech features that would be distracting in an interview prepared for broad thematic analysis. Ask the principal investigator, supervisor or client what must be retained before any cleaning or editing.
| Style | Typical treatment | When it may be appropriate |
|---|---|---|
| Full or strict verbatim | Retains defined fillers, repetitions, false starts, interruptions, pauses and relevant non-speech events | Conversation analysis, discourse studies, detailed speech or interaction analysis, or where the protocol requires it |
| Clean / intelligent verbatim | Removes specifically agreed verbal clutter while keeping substantive wording, intent, qualifiers and uncertainty | Many applied research, interview and thematic-analysis projects where readability is useful |
| Edited or publication-ready | May restructure or paraphrase wording under separate editorial instructions | Approved articles, reports and published extracts; not a substitute for the unedited research transcript |
There is no universal rule that every academic transcript must be full verbatim. Nor is the term “intelligent verbatim” interpreted identically everywhere. The specification matters more than the label. Read our full vs clean and intelligent verbatim comparison for side-by-side examples.
How to transcribe a research interview: eight practical steps
1. Confirm consent, ethics approval and the handling plan
Before recording or sending the files anywhere, confirm that collection, recording, transcription and any third-party processing are covered by the participant information, consent arrangements and relevant research ethics approvals. If the data contains personal information, follow the institution’s or commissioning organisation’s data-protection policies, approved processors and lawful handling requirements. Do not assume that participant consent for an interview automatically authorises uploading the recording into any AI service.
Decide who may access identifiable audio, who will receive the transcript, what identifiers should be removed or pseudonymised, and how long the source and final files will be retained. Sensitive recordings should not be sent through unapproved shared links.
2. Prepare complete, usable source recordings
Check each interview file’s duration and play the opening, middle and end to make sure nothing was cut off. Note the interview date, language, approximate number of speakers, file identifier and any problems such as noise, distortion, echo or missing channels. Preserve the original recording; if audio enhancement is used, review difficult words against the original rather than treating processed sound as independent evidence.
Recording setup affects every subsequent step. Use a quiet environment where possible, position microphones clearly, monitor battery and storage, and avoid relying on a distant laptop microphone for multiple speakers. See how to record quality audio for transcription.
3. Establish participant IDs, roles and a terminology list
Use the identifiers approved by the study, such as INT for interviewer and P07 for participant 7. Keep the same labels throughout the transcript. If the participant’s name appears in the audio, follow the approved procedure for storing the identifying source separately from the pseudonymised working transcript.
Prepare a short glossary for place names, organisations, acronyms and technical terms. Reference documents can help with spelling but must not override what can actually be heard. Clarify whether a bilingual interview needs an original-language transcript, a separate translation or both.
4. Transcribe speech according to the selected convention
Transcribe what is audible instead of rewriting it into more polished prose. Mark speaker changes consistently and retain substantive qualifiers like maybe, probably and I think where removing them could change a participant’s meaning. Decide whether non-speech events such as laughter or long pauses are analytically relevant, and only annotate them to the degree agreed with the researcher.
Where voices overlap, preserve the words that can genuinely be heard and mark competing speech or unclear attribution using the agreed system. Do not replace uncertain speech with a logical-looking summary. If the researcher wants a separate digest or summary, label it as an additional deliverable.
5. Add speaker labels and useful timestamps
Researchers often benefit from elapsed timestamps at speaker changes, fixed intervals or significant events. Decide the convention before work begins: for instance, a timecode such as 00:05:32 can refer to elapsed time from the start of that particular source file. If interviews are split across several files, identify the file and whether timecodes restart.
Speaker labels should reflect observable attribution, not assumptions about identity or role. A neutral Unknown speaker entry is preferable to silently misattributing a statement.
6. Handle inaudible words, interruptions and uncertainty honestly
Replay difficult sections at a suitable speed and, where practical, seek a second check from a person authorised to access the data. If speech remains impossible to identify reliably, use the client’s agreed notation such as [inaudible 00:08:41] while preserving the words surrounding it.
Use distinct labels for overlapping voices, background noise and an unintelligible single phrase only if the transcription rules define them. Do not invent words or interpret a participant’s emotional state from tone alone. A note such as [laughs] describes an audible event; [feels defensive] is a subjective inference that should not silently appear as speech.
7. Compare the transcript with the complete recording
The quality review should follow the source audio, not merely the written draft. Confirm that all sections, speaker turns and intended annotations have been captured. Pay close attention to names, numbers, dates, quoted words, negation, specialist terminology and unclear passages. A proofread can catch inconsistent spelling, but it cannot verify a missing answer or incorrect speaker without audio comparison.
Where independent review is required, record who reviewed the transcript, which files were checked, outstanding uncertainty and any corrections. The level of checking should reflect the research design and applicable quality or ethical obligations. No general percentage guarantees that an interview transcript is suitable for all studies.
8. Prepare the analysis-ready version and handover
Save the deliverable using the agreed file identifier, respondent ID, date and version. For example, IDI_P07_20261021_EN_v1.docx is a plausible filename for a fictional project, not a mandatory convention. Keep source audio, the working transcript and the approved anonymised version clearly distinguished.
Before releasing the file to the research team, confirm that any redactions or pseudonyms requested have been applied, that all source files were accounted for, and that the final format works with the intended coding or analysis workflow. Keep the access-controlled key linking research IDs to identities separate, where required by the protocol.
Example: a research interview transcript with labels and timestamps
This is a fictional illustration, not a real participant interview. It demonstrates readable, clean-verbatim formatting with one uncertain passage and a non-speech notation. Your study may require fuller notation and different timestamp rules.
| Elapsed time | Speaker | Transcript |
|---|---|---|
| 00:00:08 | INT | When did you first consider changing the service you used? |
| 00:00:14 | P07 | I think it was around March. I wasn’t unhappy with everything, but it had become difficult to get a response. |
| 00:00:26 | INT | Was the response time the main issue? |
| 00:00:29 | P07 | Probably. The cost mattered too, but I wouldn’t say it was the only reason. |
| 00:00:38 | INT | Can you describe what happened next? |
| 00:00:42 | P07 | I contacted [inaudible 00:00:45], then I decided to wait a week. [laughs] I wasn’t quite ready to change. |
The words “I think”, “probably” and “wouldn’t” remain because they may be analytically important. The transcription does not manufacture a name for the unclear passage. A longer study may additionally need interviewer prompts, pausing conventions and respondent identifiers in a header.
Research ethics, confidentiality and anonymisation
A good transcript can still expose participants if it contains names, employers, rare job roles, locations or other details that indirectly identify people. Review the entire transcript against the research protocol rather than assuming that replacing the participant’s name is enough to anonymise it.
- Consent and purpose: Ensure the recording, transcription and intended research use are covered by the study’s ethical arrangements.
- Pseudonymisation: Substitute approved identifiers such as P07 for names where required; recognise that pseudonymised data may still be personal data.
- Direct and indirect identifiers: Consider addresses, employer names, exact dates and details that may reveal a person when combined.
- Secure access: Use approved systems, restrict raw recordings to authorised people and agree retention and deletion requirements.
- External transcription and AI tools: Confirm the institution’s procurement, processing, transfer and confidentiality rules before uploading sensitive research.
- Quotations in publications: Recheck selected quotes against the source and consider disclosure risk in context.
Research ethics are project-specific. A commercial market-research study, NHS-related research and a university dissertation may have different approval processes, participant notices and data-handling conditions. The research sponsor or institution should confirm the correct framework.
From interview transcript to thematic analysis
Once transcription and quality control are complete, the researcher can import or organise the material for analysis. Typical analysis activities include familiarisation, assigning codes, comparing recurring ideas, building categories and themes, and checking interpretations against the context of the interview. These are interpretive research steps, not alterations the transcriber should make to what the participant said.
A consistent format can make downstream analysis easier: use one speaker-label convention, one approved respondent ID, agreed timecode formatting, readable paragraph breaks and consistent language treatment. Avoid replacing all spoken phrases with polished summaries, as this removes the very language researchers may need to analyse.
For the next phase, read how to code interview transcripts in qualitative research. For editing conventions and final review, see how to edit an interview transcript without changing meaning.
Special cases: market research, multilingual interviews and IDIs
In-depth interviews (IDIs) and market research
Research agencies often use IDIs to collect detailed consumer, professional or stakeholder views. A consistent transcript set helps analysts compare responses across respondents and markets. Record which interviewer prompts, respondent ID labels, brand names and product references matter to the study. Keep the original spoken claims even when an interviewee contradicts themselves; resolving that contradiction belongs in analysis.
When quoting research participants, retain the surrounding context and any agreed anonymisation. A single attractive quote should not be presented as representative of all participants without analytical support. For commissioning needs, see market-research transcription services.
Multilingual interviews
Transcribing an interview in its source language preserves the words used by the participant. Translating that transcript into English is a separate task and can affect nuance, idioms or culture-specific concepts. The research team should decide whether it needs source and translated transcripts side by side, translations of selected extracts, a terminology glossary or bilingual quality review.
Sensitive healthcare or legal interview recordings
These may involve additional confidentiality, specialist vocabulary, participant welfare or disclosure considerations. The transcriber should not invent diagnoses, interpret motives, alter admissions or substitute professional judgement for recorded speech. Ensure the relevant commissioner or researcher authorises the data handling and confirms the required degree of verbatim detail.
Interview transcription QA checklist before delivery
- Every expected interview recording opens, matches the project list and has a complete ending.
- The transcript style and handling of fillers, false starts, pauses and edited wording match the approved brief.
- Interviewer and participant IDs are applied consistently, with no guessed speaker identities.
- Numbers, names, dates, technical terms, negation and material qualifiers have been checked against audio.
- Inaudible and overlapping speech are marked according to the agreed notation rather than silently guessed.
- Timestamp references, file identifiers and sequence are consistent and verifiable.
- Any requested anonymisation or pseudonymisation has been completed under the study protocol.
- The final version is in the correct format, meets access requirements and has an identifiable review status.
For wider production controls, our 25-step transcription quality checklist provides file preparation, review, delivery and sign-off steps beyond interview-only projects.
Frequently asked questions about interview transcription
Is full verbatim always required for qualitative research?
No. Some methods require exact speech features, while clean verbatim can be appropriate for other research designs. The analysis plan and ethics-approved protocol should guide the choice. Avoid making changes that remove uncertainty or alter participant meaning.
What is the difference between transcription and coding interviews?
Transcription creates a faithful text version of spoken data. Coding assigns analytical labels to segments of that data. Coding is a separate research activity, although a carefully formatted transcript makes it easier.
How should an inaudible interview response be recorded?
Preserve the intelligible words and use a clearly documented notation such as [inaudible 00:03:52] for the uncertain part. The notation and timecode system should be agreed with the researcher. Do not fill gaps with inferred wording.
How do I anonymise an interview transcript?
Follow the study’s authorised de-identification protocol. Replace or redact direct identifiers where required, consider indirect identifiers and keep any linking key under approved access restrictions. Merely replacing the participant’s name does not always achieve anonymity.
Can AI transcribe confidential research interviews?
Only if the applicable research, information-governance, participant-consent and contractual conditions permit the selected service. Even then, the draft may need audio-following human review. Do not upload recordings to an unapproved consumer tool by default.
How long does a research interview take to transcribe?
It depends on the audio duration and clarity, number of speakers, overlap, language, verbatim style, terminology and review level. A multi-speaker or heavily detailed transcript may require much more work than clear single-speaker audio of equal length. Confirm timing after assessing the files.
Can research interview transcripts be supplied in multiple languages?
Yes, depending on the project scope. Transcription in the original language and translation into another language are distinct deliverables. Specify whether you need both versions, a bilingual layout or translated excerpts.
Need your research interviews transcribed?
For an interview transcription enquiry, supply the language, number and length of recordings, number of speakers, topic, preferred verbatim style, speaker labels, timestamps, confidentiality requirements and expected deadline. Explore professional interview transcription services or request a project quote. Availability, confidentiality arrangements and the precise deliverables should be agreed for the specific project.
