How to record an interview and get a clean transcript

Set up, then get consent on the recording itself

Start the recording first, then ask. That is the professional standard for journalists, researchers and recruiters running screening calls: their yes lands in the same file as the interview, so you never have to prove later that you asked.

The legal side varies by where each of you is sitting. Some places need only one person in the conversation to know. Others need everyone to agree, and that changes the moment one of you is in a different country or state. Read do you have to tell people you are recording and the country by country rules before an interview that could end up in print or in a hiring decision.

Before any of that, run this list while you wait. Most ruined interviews are ruined before the first question.

1. 1Start the recording and confirm it is actually running. Look for the red indicator or the moving timer, not just the button you pressed.
2. 2Say the date, your name, their name, and what the interview is for. Ten seconds. This slate also gives the transcription model a clean sample of both voices.
3. 3Ask directly: "I am recording this so I get your words right. Is that OK with you?" Wait for a spoken yes. A nod is not an answer on an audio file.
4. 4If it is on the record for publication, say so, and say what happens to the audio afterwards.
5. 5If they say no, stop the recording, say out loud that you have stopped, and take notes by hand.
6. 6If they change their mind halfway, mark the time, stop, and delete that part before you do anything else with the file.

• Charge both devices. Do not start a long interview with either one under 50 percent.
• Check storage. An hour of compressed audio is roughly 30 MB. Lossless is several times bigger.
• Turn on Airplane Mode or a Focus. An incoming call is the most common way a phone recording gets cut in half.
• Check nothing covers the microphone. A thick case over the bottom mic is a real problem.
• Do a 20-second test. Record a sentence from your seat and one from theirs, then play it back. You are listening for whether their voice is as loud as yours.

Where to put the microphone

A phone microphone is good enough for almost any interview. What decides quality is distance and what sits between the mouth and the mic, not how much you spent.

Two rules cover most of it. Closer beats better: halving the distance to the speaker does more than any app setting. And put the phone closer to them than to you, because you already know what you said. Their answers are what you will quote.

Lay the phone flat with the bottom edge, where the main mic sits, pointing at the person talking. Rest it on something soft: a folded napkin, a notebook, a scarf. Bare wood and glass carry every knock and every cup being set down straight into the recording.

Hard cases: a noisy cafe, a phone call, a video call

The cafe. Steady noise is the enemy, not loud noise. One dropped tray is a second you can work around. An air conditioning unit humming under every answer sits under the whole hour and cannot be fixed afterwards. So walk in five minutes early and pick the table: away from the counter, the door, the speakers and any vent you can hear. A corner or a booth beats the middle of the room, because two walls block half the noise. Sitting at ninety degrees, on the corner of a table, beats sitting straight across, because you both end up closer to the phone. A clip-on lavalier microphone costs very little and removes the problem. Clip it a hand's width below the chin, clear of scarves and collars.

The phone interview. Do not assume your handset can record the call. On iPhone, iOS 18.1 and later can record from the Phone app in supported regions: tap the record button, iOS announces it to both sides, and the file lands in the Notes app. On Android it depends on the brand and the country, and in many countries the feature is switched off. The method that works everywhere is lower tech. Put the call on speakerphone on one device and record it with a second device next to it. Quality drops a little, reliability goes up a lot. More in how to record a phone call on iPhone.

The video call. Either use the platform's own recording, which needs the host to allow it and warns everyone, which is a feature for interviews rather than a problem. Or send an AI note taker that joins as a participant. Noter AI does that on Zoom, Google Meet, Microsoft Teams and Webex: paste the link, or connect Google Calendar or Outlook and let it join scheduled calls. Everyone sees it in the participant list.

Record the interview on iPhone

The free built-in route is Voice Memos. It works on any iPhone and needs nothing installed.

1. 1Open Settings, tap Voice Memos, set Audio Quality to Lossless if you have the storage. Compressed is fine for a normal room.
2. 2Swipe into Control Center and turn on Airplane Mode.
3. 3Open Voice Memos and tap the red record button.
4. 4Place the phone as described above, then say your slate and ask for consent.
5. 5Leave the phone alone. You can lock the screen and the recording keeps going. Do not pick it up mid-answer.
6. 6Tap stop at the end, then rename the file straight away. "New Recording 47" is how interviews get lost.
7. 7To read it on the phone, open the recording and tap the transcript button, on iOS 18 and later. Free and quick, but plain text with no speaker labels, no clickable timestamps, one language at a time.
8. 8For a transcript you can work from, tap the three dot menu, tap Share, and send the file to a transcription app. The same share sheet works from WhatsApp voice notes and third party recorders. See how to transcribe a voice memo on iPhone.
9. 9Or skip the handover: record inside Noter AI instead, where New Note then Start Recording keeps running with the screen locked and a Control Center widget starts one in a single tap. Transcription begins the moment you stop.

Record the interview on Android

The built-in recorder is different on every brand, so start by finding yours.

1. 1On a Pixel, open the Recorder app and tap the red record button. It transcribes on the phone as it records and lets you search inside recordings.
2. 2On a Samsung, open Voice Recorder, tap Recording mode, choose Interview. That mode uses the top and bottom microphones separately for two people facing each other, so lay the phone flat between you with one end pointing at each of you.
3. 3On any other brand, use the recorder that came with the phone or install one from Google Play. Set the format to M4A or WAV, not a low bitrate option.
4. 4Turn on Airplane mode from quick settings before you start.
5. 5Start recording, place the phone, then say your slate and ask for consent.
6. 6Open your battery settings and set the recorder app to Unrestricted if that option exists. Aggressive battery saving kills background recordings.
7. 7Tap stop at the end and rename the file immediately.
8. 8For a transcript with speaker labels, record the next one directly in Noter AI on Google Play so nothing has to move between apps.

Record on two devices, always

For anything you cannot repeat, record twice. An interview you travelled to, a source who agreed once, a candidate about to fly home, a researcher's only session with a participant. Two devices turn a disaster into an inconvenience.

The backup does not need to be good. An old phone at the far side of the table, a laptop, a cheap dictaphone in your bag. If you are thinking of buying a dedicated recorder for this, check first whether a Plaud-style device beats the phone you already carry. What matters is that it fails in a different way than the first one. Two apps on the same phone is not a backup, because one dead battery takes both. There is a bonus: a second phone at the other end of the table puts a mic near each person, so if one voice comes back weak on the main file, the backup often has it clean.

• Start the backup first and stop it last. It should always be the longer file.
• Check both after 30 seconds. A device that never started looks a lot like a device that is recording.
• Different power sources, so one flat battery cannot end both.
• Different distances, so a knocked cup does not land on both files.
• Copy both files off the devices the same day, before you edit, travel, or drop the phone.
• Keep the raw audio until the work is published or the file is closed. A transcript is not evidence. The recording is.

Interviews in another language, or in two at once

This is where most transcription tools quietly fall apart, and it is worth knowing before you rely on one in the field.

Many apps make you choose the language before you press record. Fine for an interview held entirely in French. It breaks the moment your subject does what bilingual people naturally do, which is answer in one language and reach for another for the words that live there. A founder in Dubai answers in Arabic but says runway, term sheet and cap table in English. A doctor in Mumbai explains in Hindi and names every drug in English. A tool locked to one language either writes those words phonetically in the wrong alphabet or drops them. Either way, the quotable part of the answer is the part that got destroyed.

The fix is a model that tags language word by word instead of committing to one language per file. Noter AI works that way. Transcription covers 60+ languages with detection always on, so there is no language to select and nothing to get wrong when a stranger starts talking, and a sentence carrying two languages arrives with both halves intact.

Finished notes can then be translated into 18 languages: Arabic, Chinese (Simplified), Chinese (Traditional), Dutch, English, French, German, Indonesian, Italian, Japanese, Korean, Polish, Portuguese, Russian, Spanish, Thai, Turkish and Vietnamese. So you keep the original words for the quote and hand a translated summary to an editor who does not read the source language. More in the multilingual transcription roundup and, if it applies to your work, the Arabic interview page.

A transcript you can quote from

A wall of unbroken text is not something you can work with. Four things separate a transcript you skim once from one you can build an article, a report or a hiring decision on.

Speaker labels tell you who said what. With two people and a well placed mic this is usually clean. Crosstalk and similar sounding voices are what confuse it. See can AI tell who said what.

Timestamps on every line make a quote checkable. Find the sentence, tap it, hear the exact moment. Ten seconds of listening is the difference between an accurate quote and a correction.

An editable transcript matters more for interviews than for meetings, because interviews are full of names, places and technical terms no model has seen. Fix them once in the text, and in an app that reruns the summary afterwards the correction flows into the summary and the action items too. See editing a transcript and rerunning the summary.

Search across everything turns a year of interviews into something useful: fifteen conversations on the same subject, one question, an answer drawn from all of them.

• Read the transcript once with the audio playing before you quote from it.
• Mark good quotes with their timestamp as you go, instead of hunting at the end.
• Keep the speaker's own words, and say so if you tightened a quote.
• Export to PDF for a fixed record, Word if an editor marks it up, plain text for pasting anywhere.

Troubleshooting: five things that go wrong

• One voice is much quieter than the other. Almost always placement: the phone sat closer to you. On the file you have, raise the level on the quiet passages in a free audio editor and re-transcribe, or check the backup device at the other end of the table, which may already be clean.
• The speaker labels are wrong, or two people got merged. This happens with similar voices, crosstalk, or a mic too far from everyone. Fix the labels once and give each speaker a real name, then reread the fast, overlapping sections, because that is where swaps cluster. The ten second slate at the start prevents most of it.
• Long pauses made you think it stopped. Silence is not an error, and twenty seconds of thinking before a hard answer is often the best moment in the interview. Do not stop and restart to save space, or you end up with fragments and a gap exactly where they started talking again.
• The recording stopped part way. Usual causes: an incoming call, storage filling up, or battery saver killing the app in the background. Airplane Mode, a storage check and Unrestricted battery use on Android prevent all three. If it already happened, this is what the second device was for.
• The interview ran three hours. Do not split it while it is happening, because one long file is easier to handle than five fragments. Afterwards, do not read 25,000 words. Read the summary, then use search and the timestamps to jump to what matters, and check the action items for anything you promised the person. Expect a long file to take longer to come back.

If you would rather not type it yourself

Typing an interview takes roughly four to six hours per hour of audio, and it is the least valuable part of the job. Everything above works with any recorder and any transcription tool. If you want it all in one place, that is what Noter AI is for.

Record from the phone on the table, send the bot into a Zoom, Meet, Teams or Webex interview, or import audio you captured elsewhere. What comes back is a transcript with speaker labels and timestamps, tap any line to hear it, an editable transcript that reruns the summary once you fix a name, action items pulled out on their own, and AI chat that answers questions across one interview or across every interview you have recorded.

It is $9.99 a month or $49.99 a year, one flat price rather than per seat, with no cap on how many interviews you run or how many summaries you generate. Interview work comes in bursts, which is where metered plans hurt: a week of back-to-back fieldwork is exactly when a monthly credit allowance runs out. There is a free trial, and the honest test is one real interview, not a rehearsal.

Get it on iPhone and iPad or on Android. One limit stated plainly: it takes audio, not video, so if your interview exists only as a video, pull the audio track out first.