How to transcribe interviews for qualitative research

The short answer

Decide how you'll handle participant data first, then transcribe with speaker labels and timestamps, proofread against the audio, and export to a format your analysis tool imports. Automatic transcription on your own computer saves most of the typing and keeps recordings inside your approved storage; the proofreading pass is still yours.

Before you record: consent and data handling

Your consent form, and your ethics board or data protection plan if you have one, usually say where recordings may be stored and who may process them. Many cloud transcription services count as a third party processing participant data, which may need to be disclosed or approved. A tool that runs on your own machine avoids that question, but check your protocol either way.

Plan pseudonyms and a de-identification step before you share transcripts with collaborators. Automatic tools don't anonymize for you.

Verbatim or clean?

• Verbatim keeps fillers, false starts, pauses and overlaps. Useful for discourse or conversation analysis.
• Clean (intelligent) verbatim drops the ums and repairs, keeping meaning. Most thematic analysis works fine with it.
• Automatic transcripts tend to land in between. Decide which you need and adjust in proofreading, noting the convention in your methods section.

How to do it in MediaFind

MediaFind transcribes on your computer, with no upload, account or API key.

1. Put the interview recordings in one folder and add it to MediaFind.
2. Add project terms and names to the names & terms list so they're spelled consistently.
3. Let it transcribe. Each interview gets speaker labels and word-level timestamps; name the interviewer's voice once and it's recognized in the other interviews.
4. Proofread against the audio: click any line to play from that moment.
5. Export each transcript as Word or text for your qualitative analysis software, or as SRT or VTT if you need timestamps in the file.
6. Use search and Ask across all interviews for a first look at themes, e.g. moments where participants describe feeling rushed. Answers cite the interview and timestamp.

A proofreading routine that scales

Proofreading takes the most time, so make it systematic. Listen at a slightly faster speed with the transcript open, fix names and terms first with find and replace, then correct speaker labels, and finally mark inaudible passages with a consistent tag and timestamp rather than guessing.

Keep a short log per interview of what was corrected and any passages you couldn't resolve. It helps with inter-coder discussions later and makes your methods section easier to write.

Tips and limitations

Treat semantic search and Ask as a way to find passages, not as coding: they help you locate evidence faster, but the analysis and the codebook are yours. Keep an unedited copy of each automatic transcript so you can show what changed during proofreading.

Accents, crosstalk and phone-quality audio all raise the error rate, so budget more proofreading for focus groups than one-to-one interviews. The free plan covers 10 recordings, enough for a pilot. See researchers and UX research interviews, and identifying speakers.

Frequently asked questions

Is automatic transcription good enough for qualitative research?

As a first draft, usually yes. Proofread against the audio before coding or quoting, especially for accents, jargon and overlapping speech.

Can I transcribe interviews without uploading participant data?

Yes. MediaFind transcribes on your computer and uploads nothing, which can make ethics and data-protection approval simpler. Check your own protocol.

Which export format should I use for analysis software?

Word or plain text is the most widely accepted. MediaFind exports Word, PDF, SRT, VTT and text.

Does MediaFind anonymize transcripts?

No. Replace names and identifying details yourself before sharing transcripts outside your research team.