Volver al Blog

How to Record an Interview (So It's Actually Usable Afterwards)

How to Record an Interview (So It's Actually Usable Afterwards)

Most bad interview transcripts are not a transcription problem. They are a recording problem that shows up later.

By the time you upload the file, the decisions that determine whether you get clean, speaker-separated text have already been made — where the microphone sat, how the room sounded, whether two people talked over each other. None of that can be fixed afterwards.

Here is what actually matters, in rough order of impact.

1. Distance beats equipment

If you change one thing, change this. The distance between the microphone and the person speaking matters more than what the microphone costs.

Sound falls off quickly with distance while room noise stays constant, so a cheap microphone close to someone's mouth will usually beat an expensive one across a table. A phone lying 30cm from the speaker in a quiet room produces a genuinely good recording. The same phone two metres away in the same room produces something you will regret.

Practical version: put the recording device on the table in front of the person talking, not equidistant between you both.

2. One microphone per person, if you can manage it

The single most common interview setup — one phone in the middle of the table between two people — is also the hardest case for automatic speaker separation.

Both voices arrive at similar volume, from similar directions, with similar room reflections. Those are precisely the cues that distinguish speakers. Remove them and even a good transcription tool has little to work with.

You do not need professional gear to fix this. Two phones, one in front of each person, each recording separately, is a large improvement over one phone in the middle. If you only have one device, place it closer to your subject than to yourself — you know what you said; it is their words you need.

3. Fix the room, not the audio

Thirty seconds of attention before you start beats an hour of cleanup afterwards:

  • Hard rooms echo. Bare walls, glass, and big tables reflect sound. Soft furnishings, curtains, and carpet absorb it. Given a choice of rooms, take the softer one.
  • Find the hum. Air conditioning, a fridge, a projector fan, traffic through an open window. You stop noticing these within a minute; the microphone never does.
  • Nothing on the table you'll touch. Tapping the table, shuffling papers, or setting down a cup right next to the microphone produces sharp noises that transcription handles badly.
  • Phones flat and silenced. A phone buzzing on a hard surface next to the recorder is the most avoidable ruined take there is.

4. Match the setup to the interview type

In person

Voice recorder app on a phone is fine. Placement as above. If it matters and you can, run a second device as a backup.

Over the phone

Ordinary phone calls are the awkward case, because whether you can record depends on your handset, carrier, and country. On Android, that means your phone's built-in dialer recorder rather than a third-party app — we cover the current state of that in how to record a call on Android.

The fallback that always works: put the call on speaker and record with a second device. Audio quality drops and both voices arrive through the same speaker, which makes speaker separation harder — but you get the interview.

Remote video call

Use the platform's own recording where you have it. Recordings made this way often keep each participant's audio cleanly, which is exactly what you want. If you are recording a Zoom or Teams call, the file you get can go straight into transcription without extracting the audio first.

5. Ask permission, on the recording

Ask before you start, then start the recording and ask again so the answer is captured in the file itself. It takes three seconds and it documents the consent given at that point — good practice rather than a guarantee, since consent can be limited or later withdrawn.

The legal position varies: some jurisdictions require only one party to consent, others require everyone, and covert recording can carry civil and/or criminal consequences. If your interview is for academic research, journalism, or anything that will be published, your institution or publisher will usually have consent requirements of its own on top of the law. This is general guidance, not legal advice — check what applies where you and your subject are.

6. Say who is speaking, at the start

A trick that costs nothing and saves real time later: open the recording with each person saying their name. "I'm Sarah, I'm interviewing Dr Patel."

Automatic transcription groups voices under generic labels — Speaker 1, Speaker 2 — because it has no way to know their names or roles. If the first thirty seconds of the recording establishes who is who, relabeling the transcript afterwards takes seconds instead of requiring you to hunt for a distinguishing line.

7. Record a backup for anything you cannot repeat

The failure modes are boring and common: storage fills up, an app is interrupted by an incoming call, a battery dies, a file will not open.

If the interview can be redone at no cost, accept the risk. If your subject travelled to meet you, is being paid, or is only available once, run a second device in parallel. It costs nothing.

Afterwards: turn it into text

Once you have the file, the interview becomes far more useful as text than as audio. You can search it, quote from it accurately, skim an hour in a couple of minutes, and code it for themes.

Upload the recording to our interview transcription tool — each distinct voice comes back grouped under its own editable label, so you can see where the conversation changes hands and name the speakers yourself. Recordings up to five minutes are free with no signup.

If you want the transcription side in more depth — editing, speaker labels, exporting for analysis — see how to transcribe an interview.

The one-minute checklist

  • Microphone close to the speaker, not in the middle of the table
  • Quiet room, nothing humming, nothing on the table you will touch
  • Ask permission, with the recording running
  • Everyone says their name at the start
  • Backup device for anything you cannot repeat
  • Check the first minute plays back before you rely on it