How to transcribe an interview
Getting a usable interview transcript is mostly about what you do before anyone starts talking.
· 6 min read
Transcription is the least glamorous part of interviewing and the part most likely to cost you a quote. The good news is that most of the work happens before the interview starts: get the recording right and the transcript mostly takes care of itself.
Here is the sequence that works, whether you are transcribing by hand, with software, or live.
Ask permission, and record that you asked
Recording laws vary by country and by state, and "everyone knew I had a recorder out" is not the same as consent. The safe habit is to ask on the record, at the start, after the recording has begun — so the permission is captured in the same file as the interview.
A single sentence does it: "I am recording this so I can quote you accurately — is that alright?" It takes five seconds and it settles the question permanently.
Put the microphone closer than feels necessary
This is the single highest-leverage thing you can do, and almost everyone gets it wrong. Sound falls off fast with distance — roughly, doubling the distance quarters the energy reaching the microphone. A phone at arm's length across a café table is picking up far more room than voice.
Practical version: put the recording device on the table between you, closer to your subject than to yourself, off any surface that rattles, and away from the wall. If you are using a laptop, be aware its microphone is usually aimed at the screen hinge, not at the person opposite.
If you can only fix one thing about your setup, fix this one.
Deal with the room before you deal with the software
No transcription tool recovers information that never reached the microphone. A few seconds of attention pays off:
- Choose a corner or a booth rather than the middle of a room — fewer reflective surfaces
- Sit away from air conditioning, fans and fridges, which produce steady noise that masks speech
- Avoid a table by the door or the coffee machine
- Soft surroundings are better than hard ones: curtains and upholstery absorb, glass and concrete reflect
Transcribe live, or transcribe after
There are two workable approaches, and the right one depends on what you need the transcript for.
Transcribing live means the text appears as the conversation happens. The advantage is not really speed — it is that you can see what was captured while you can still do something about it. If a name comes out garbled you can ask them to spell it there and then, rather than discovering the problem three days later with no way to check.
Transcribing afterwards from a file lets you use slower, more accurate processing, and it is the right choice when you already have a recording, or when you need the highest possible accuracy and can wait for it.
Clean up in the right order
Resist the urge to fix everything top to bottom. Work in passes, cheapest first:
- Fix speaker attribution before wording — a misattributed quote is a much worse error than a misspelled one
- Search for proper nouns, company names and numbers, and check those against the audio
- Only then read for sense and tidy the wording
Most transcription errors cluster around names, jargon and figures — exactly the things you are most likely to quote and least able to guess from context.
Quote from the audio, not the transcript
Before a direct quote goes to print, listen to it. Every transcript, human or automatic, is an interpretation, and the sentence you most want to use is often the one delivered fastest and least clearly.
This is why word-level timestamps matter more than they sound: being able to jump straight to the moment a sentence was spoken turns verification from a chore into a few seconds of work.