How to transcribe an interview

Transcribing an interview is a higher-stakes version of transcribing anything else. You're not just trying to remember what was said — you're probably going to quote it directly, which means the bar for accuracy on names and specific phrasing is a lot higher than for a transcript you'll only skim.

The single biggest thing that wrecks accuracy

It's not background noise. It's overlapping speech — the interviewer jumping in with a follow-up before the interviewee finishes a sentence. A transcription model has to guess which voice to follow during an overlap, and it often guesses wrong for the rest of that sentence, not just the overlapping words. If you're the one conducting the interview, the highest-leverage habit is letting people finish before you respond — it costs nothing in the moment and meaningfully improves the transcript later.

Names and jargon are the other predictable failure point, and no amount of audio quality fixes that — the model is guessing at a word it's never encountered, the same way a person would mishear an unfamiliar name over the phone. Budget a proofreading pass specifically for proper nouns before you quote anything.

Interview Transcription accepts the recording as audio or video directly — a phone recording, a Zoom call export, or a dedicated voice recorder file all work without conversion first.