How to Transcribe Interviews: A Guide for Journalists & Researchers

Why Interview Transcription is Different

Transcribing an interview isn’t like transcribing a podcast or a meeting. The stakes are higher:

• For journalists, a misheard quote can become a correction, a retraction, or a defamation lawsuit.
• For qualitative researchers, transcripts are primary source data — they feed coding, thematic analysis, and published findings that other scholars will cite.
• For oral historians, the transcript may outlive the interviewee and become the canonical record of a life or an event.

That means three things matter more than speed: accuracy, verbatim fidelity, and confidentiality.

The Classic Problem: Time and Privacy

Manual transcription of a single hour of interview audio takes 4–6 hours for an experienced transcriber. It’s why services like Rev, Trint, and Otter have thrived — they trade money for time.

But every cloud transcription service creates the same dilemma: your source’s audio is now sitting on someone else’s server, often protected only by terms of service and an aspirational privacy page. For confidential sources, vulnerable populations, IRB-approved research, or legally sensitive interviews, that’s a risk you can’t always accept.

Browser-based AI transcription changes the math. A 60-minute interview can be transcribed in under 10 minutes, on your laptop, with the audio never leaving your device.

Before the Interview: Recording Well

Transcription accuracy is determined at the recording stage. The best AI in the world can’t rescue bad audio.

Equipment

• A dedicated recorder or mid-range lavalier mic always beats a phone’s built-in mic.
• Two mics are much better than one. A separate track per speaker eliminates crosstalk and makes speaker separation trivial.
• For remote interviews: have the subject record their side locally with a free tool, and send you the file. Zoom/Teams audio compresses noticeably.

Environment

• Hard surfaces and bare walls cause echo. Record in furnished rooms when possible.
• Close windows, turn off HVAC when you can do so politely, and silence phones.
• Avoid cafés for anything sensitive — ambient chatter is the hardest kind of noise for AI to filter.

Sample Rate & Format

• WAV or FLAC at 44.1 or 48 kHz is ideal if you’re archiving.
• MP3 at 128–192 kbps is fine for AI transcription and saves huge amounts of storage for long interviews.

The Transcription Workflow

1. Back up the raw audio first

Before you do anything else, copy the audio file to a second location. Losing a Pulitzer-winning quote to a dead SD card is a career event.

2. Transcribe in the browser

Open a local tool like Whisper STT. Upload the file. Choose the Small or Medium model — for interview accuracy, it’s worth the extra minute of runtime compared to Tiny.

Whisper does remarkably well with:

• Mixed accents and non-native English speakers
• Soft-spoken or emotional speech
• Low-volume passages (as long as they’re audible)
• Technical jargon (with a glossary pass afterward)

3. Choose your transcription style

Researchers and journalists need different things:

Style	What It Captures	Use For
Clean / edited	Meaning, minus filler	News articles, op-eds, feature writing
Verbatim	Every word, including “ums”	Q&A articles, direct quotes
Full verbatim	Words + pauses + non-verbal cues	Qualitative research, oral history

Whisper produces a clean-ish transcript by default. For full verbatim, you’ll do a pass listening back and inserting markers like [laughs], [long pause], or [crosstalk] where they matter.

4. Verify every quote you plan to publish

Non-negotiable. Before any direct quote leaves your keyboard, play the exact audio segment at 0.75x speed and confirm the words match. AI transcription is excellent but not infallible, and the transcript is your draft, not your ground truth. The audio is.

5. Tag, code, and archive

For qualitative research, export to plain text and import into NVivo, ATLAS.ti, MAXQDA, or Dedoose. For journalism, drop it into your CMS, Google Doc, or investigative notes system — with an obvious link back to the timestamped audio.

Ethics and Confidentiality

A few rules the best interviewers live by:

• Tell sources you are recording. Many jurisdictions require it; ethics require it everywhere.
• Explain where the audio will be stored and who can access it. “On my laptop, transcribed by AI that runs on my laptop, never uploaded anywhere” is a clean answer you can give in good faith with a local tool.
• Honor off-the-record agreements. If a segment is off the record, delete it from the transcript before sharing it with editors.
• Secure the archive. Full-disk encryption on your machine, plus an encrypted backup, is the baseline for any journalist or researcher working with source material.

Academic Context: IRB and Data Protection

For research covered by an Institutional Review Board (IRB) or GDPR:

• Local transcription is often easier to approve than cloud transcription — there’s no third-party processor to disclose.
• Document the workflow in your data management plan: file locations, retention period, anonymization steps, destruction date.
• If participants are identifiable in the audio (accent, name, professional role), plan for an anonymization pass on the transcript before it’s shared with co-investigators.

Getting Started

Get hours of your week back without compromising your sources. Transcribe interviews privately with Whisper STT — in-browser, open-source AI, no uploads, no account.