Key Takeaways
Record on the phone in your pocket, in a quiet room, about a foot from the person talking. Room noise ruins amateur recordings; microphone quality rarely does. Ask open questions – what a street looked like, what a job involved day to day, how two people met – and stop after twenty minutes. Then keep three copies on two kinds of storage with one of them somewhere else, write who, when and where into every filename, and make a transcript as well as keeping the audio.
There is nothing to buy first. The phone you already carry, used carefully in a quiet room, will produce a recording that outlasts you, and it will do it today rather than after a weekend of research into microphones. The hard parts of this are not technical. They are asking questions that get a real answer, keeping the session short enough that the person enjoys it, and then looking after the file so it is still openable and still findable in twenty years.
If you have arrived here with a folder of recordings of somebody who has already died, skip ahead. The storage and labelling advice applies to what you have exactly as it applies to what you are about to make, and it is the half that gets neglected.
A quiet room beats good equipment
What wrecks amateur recordings is the room, not the microphone. A phone held about a foot from the speaker in a quiet room will sound better than an expensive setup used badly in a kitchen with the extractor fan running. Before you press record, listen to the room for a moment with the intent of hearing everything except the person. Traffic, a television two rooms away, a fan, a fridge that cycles on and off. Turn off what you can turn off and shut the door on the rest.
Then put the phone somewhere it will stay put, roughly a foot away, and leave it alone. Don’t hold it in your hand and gesture with it. Don’t rest it on the arm of a chair that someone will drum their fingers on. Check the first thirty seconds by playing them back before you commit to a whole session, because a recording that turns out to be unusable is worth catching in the first minute rather than the last.
If they live far away
Recording a phone call is more awkward than recording a person in a chair, and the rules differ – what is legal varies by where you both are, and carriers and handsets differ in what they offer. Ask permission first, on the recording, and get the answer on the recording too. The most reliable low-tech route is to put the call on speaker in a quiet room and record it on a second device. Their side will sound thinner than yours. That is a real loss and it is still worth having.
Ask the questions that cannot be answered with yes
The difference between a recording that gets replayed and one that sits unopened is almost entirely in the questions. Closed questions produce closed answers. “Did you like living there?” gets you a yes. “What did the street look like when you walked out of the front door?” gets you a description, and inside the description there is usually a person, a shop, a smell, a job someone did that no longer exists.
Aim for questions that require a picture or a sequence to answer:
- What did the street look like where you grew up?
- What did your job actually involve, hour by hour, on an ordinary day?
- How did you and Mum meet – what happened first?
- Who else lived in the house, and where did everyone sleep?
- What did you do on a day when there was nothing to do?
Then get out of the way. The most useful thing you can say is nothing, followed by “and then what?” People stop talking because the listener starts talking, not because they have run out. Resist correcting dates. If they say 1963 and you are certain it was 1964, let it stand; you are recording their account, and you can annotate it later.
Twenty minutes, then stop
Twenty minutes is a realistic session. Long sessions tire everyone, and the second half is almost always worse than the first – answers get shorter, the voice gets flatter, and the person starts to feel examined rather than listened to. Ending while it is still going well means there is a next time, and the next time will start faster because you both know how it works.
Three short sessions on three visits will give you more usable material than one long afternoon, and they let you follow up. You will listen back and find the thing you wish you had asked about. Write it at the top of the list for next time.
Name the file before you do anything else
Labelling matters more than any technical decision in this article. A folder of files called voice memo 14 is a folder nobody will ever open twice. Write who is speaking, when it was recorded and where, into the filename itself, not into a note somewhere else that will get separated from the audio. Whoever opens the folder in twenty years will not recognise the voice and will not know which grandmother it is.
On format: uncompressed WAV keeps everything and the files are large. For speech, a high-bitrate M4A or MP3 is a reasonable preservation copy and a great deal easier to store, back up and email to a cousin. If your phone gives you the choice and storage is not tight, WAV does no harm. If it doesn’t give you the choice, don’t lose sleep over it – the recording you actually made beats the format you didn’t.
Three copies, two kinds of storage, one somewhere else
One copy is not an archive. The rule archivists use is three copies, on two kinds of storage, with one of them in a different place. That sounds elaborate until you count what you probably have now: a single copy on a phone that lives in a pocket near stairs, water and thieves.
A workable version for a family: the files on your computer, a copy on an external drive kept at another address or given to a sibling, and a copy in cloud storage. Cloud storage counts as one of the three, not as a backup by itself. Accounts get closed, terms change, and shared cloud folders have a habit of vanishing when the account holder dies – which is precisely the scenario this whole exercise exists for. Give a relative their own copy on their own drive, under their own control.
Keep the original untouched. Whatever editing, trimming or cleaning you do, do it on a duplicate with a different filename. Original recordings are the thing that cannot be remade.
The transcript does a different job from the audio
You want both, for different reasons. The audio is the voice – the accent, the pauses, the laugh, the thing a grandchild will play to hear what she sounded like. A transcript will never replace that and shouldn’t try. What it does is make the recordings usable: searchable by name and place, quotable in a letter or a eulogy, and readable by relatives who will never sit down for two hours of listening but will happily read four pages.
You can type it yourself, which is slow but forces you to hear every word, or run it through software and correct the result. If you go the software route, SozAI is a transcription app for iOS, Android and macOS that produces text with speaker labels and summaries in a long list of languages, which helps when two people are talking over each other. There is also a walkthrough of transcribing audio to text if you want the general method before you pick a tool.
Transcribing everything is usually the wrong move. Do the recordings you would be sorry to lose the content of, keep the rest as audio with good filenames, and come back to them when someone asks.
What you should expect to go wrong
Older voices, strong accents, dentures, hearing aids and two people finishing each other’s sentences are all hard for automatic transcription, and any audio to text tool will hand you a draft rather than a finished document. Names of villages and dead relatives are where it will fail most often, and those are exactly the words you care about. Budget an hour of correcting for material you intend to keep as text.
Cleaning up audio has limits too. Noise reduction can lift a voice out of a hissy recording, and pushed too far it makes the voice sound like it is coming through a wall. Hum and hiss can be reduced; a television in the same room cannot be removed. This is the argument for the quiet room, made after the fact.
And expect the conversation itself to be uneven. Some sessions produce twenty minutes of gold and some produce a polite summary of things you already knew. That is normal, and it is not a reason to stop. The recording you make on an ordinary Tuesday, badly lit, with the dog in it, is the one that will be played.
