Guide

How to transcribe an interview with speaker labels

Updated October 2026 · 28 October 2026

To transcribe an interview with speaker labels, record it with the person's consent on a phone placed between you, run it through a transcription app that does speaker labels (diarization), and rename "Speaker 1" and "Speaker 2" to names or participant IDs. Then check every passage you'll quote against the audio, because names, numbers and crosstalk are where transcripts go wrong. Finally, pick verbatim or clean verbatim, add timestamps, and store the file where your consent form says you will.

Consent comes first

Ask before you start, then ask again once the recording is running: "You're happy for me to record this?" A yes on the audio is the cleanest record of consent there is. Journalists and researchers often go further: researchers follow the consent form their ethics board approved (including where recordings and transcripts may be stored and who may process them), and reporters follow their newsroom's rules.

For phone and video interviews in the US, the rules depend on the states involved. Federal law lets one party to a conversation record it, but around a dozen states require everyone's agreement in some or all situations, and when callers are in different states the safe assumption is the stricter law. We summarise the US and UK basics in our consent overview.

The steps

1. Get consent on the recordAsk before you start, then ask again once recording so the agreement is in the audio. For research, follow your ethics approval's consent form.
2. Record close to both voicesPut the phone between you, flat on the table, in a quiet room. Avoid cafés: background chatter is the hardest thing for transcription.
3. Use an app with speaker labelsSpeaker labels (diarization) mark each voice as Speaker 1, Speaker 2. Introduce yourselves by name at the start to make naming them easy.
4. Rename the speakersSwap Speaker 1 and Speaker 2 for names, or for participant IDs if the interview must be anonymous.
5. Check every quote against the audioListen to each passage you'll quote. Names, numbers and overlapping speech are where transcripts go wrong.
6. Format and store it safelyChoose verbatim or clean verbatim, add timestamps, remove identifying details if needed, and keep the file where your consent form says it will be kept.

Recording an interview that transcribes well

• Choose the room. A quiet office beats a nice café. Background conversations are the hardest thing for any transcription engine, and they confuse speaker labels too.
• Keep the phone equidistant. If the phone sits right by you, your questions will be crisp and the answers faint.
• Don't talk over the answer. Hold your "mm-hm" until they've finished. Overlapping speech is where labels swap.
• Say names and spellings. "That's Siobhan, S-I-O-B-H-A-N." No transcript will guess an unusual name from sound alone.
• Remote interviews: the call platform's own recording usually sounds better than a phone listening to a laptop speaker. If you record a call through a laptop speaker instead, keep the volume comfortable and the phone close to the speaker, and tell the other person.

What speaker labels can and can't do

Speaker labels (the technical term is speaker diarization) cluster the recording by voice and mark each turn: Speaker 1, Speaker 2. With two people in a quiet room they're usually reliable. They get shaky with similar voices, very short turns ("yes", "right"), people talking over each other, and recordings that were paused and resumed. Some apps improve labels with a second pass over the whole file once you stop.

Labels tell you that voices differ, not whose they are. Naming them is your job, though some apps help: if someone says their name on the recording, the app can suggest it.

Verbatim or clean verbatim?

Verbatim keeps every filler, false start and pause. It's for when how something was said is part of the data, as in conversation analysis or some qualitative research. Clean verbatim removes the "ums", repetitions and stumbles while keeping the meaning and wording; it's the norm for journalism, user research and reports. Automatic transcripts usually land somewhere in between, so decide which you need and edit consistently.

Timestamps, anonymity and storage

• Timestamps at each change of speaker let you jump to the audio to check a quote.
• Anonymity: replace names with participant IDs (P01, P02), and remove details that identify someone indirectly: employer, town, a distinctive job title.
• Storage: keep audio and transcripts where your consent form or newsroom policy says, and delete the audio when you no longer need it.

APA tips for interview transcripts

If you're writing up research in APA 7: quotes from participants in your own study are part of your data, so you present them in the text (with a participant ID) and don't list them in the references. An interview you did for background rather than as research data counts as a personal communication: you cite it in the text only, for example (T. Nguyen, personal communication, October 14, 2026). Check your department's requirements for locators and longer quotations.

Where Fenwrite fits

Fenwrite is our app. It labels each voice while the interview happens, suggests names when people introduce themselves, and lets you rename speakers (or merge two labels that turned out to be one person). A paused or interrupted recording gets a second pass over the whole file for consistent speakers. You can copy or share the transcript as plain text with names and timestamps on every turn, plus a summary of the key points.

Before you use it for research: the audio is sent to our speech-to-text provider (AssemblyAI) to be transcribed, and the transcript's text to OpenAI for the summary, so check that your ethics approval allows that. It records in the app only and can't import an existing file. The free hour allows 10 minutes a recording; Pro allows 3 hours. Fenwrite is coming soon to the App Store (iPhone and Apple Watch) and Google Play (Android).

Questions

Keep reading

AI note taker without a botRecord in the room instead of sending a bot, plus US and UK consent basics. Record and transcribe a meeting on iPhoneVoice Memos, Notes, or an AI note taker with speaker labels: steps and tips. Best Otter alternatives (2026)Eight picks by reason for leaving: price, bots, languages, hardware, free.