Record Your Child's Voice Once a Month

The average parent has several thousand photographs of their child and no recording of what they sounded like at eighteen months.

Nobody decided this. It is just what the camera app makes easy. The shutter button is the first thing you see; the voice recorder is three taps away in a folder you never open. So we have documented our children’s faces exhaustively and their voices almost not at all, and most people do not notice until the voice has already changed.

A voice dates a childhood more precisely than a face

Look at a photograph of a child at two and the same child at three. It is recognisably the same person. The face has stretched a little, the hair is different, but you would pick them out of a line-up without hesitating.

Now listen to the same two years. They are not the same person. At two the pitch is higher, the consonants are approximate, the rhythm is uneven because the sentence is being assembled in real time. At three there is fluency, and intonation that is doing work, and — this is the part that gets people — the first traces of the accent they will keep for life. A face at two and at three is a continuum. A voice at two and at three is two different children.

Which means a voice recording pins a moment far more exactly than a photograph does. You can look at a picture and think, roughly, she must have been two and a half. You hear thirty seconds of audio and you know, immediately and without effort, precisely where you were.

What actually disappears

It is worth being specific about what is being lost, because “their voice” is vague enough to feel replaceable, and it is not.

The pronunciation errors. Every family has them and every family swears they will never forget them. They are corrected out of existence within a few months and they do not come back, because once the child says the word properly everyone around them adopts the correct version instantly.

The pitch. It drops steadily and you cannot reconstruct it from memory. People are extremely bad at remembering pitch — much worse than they are at remembering faces — which is why hearing an old recording is so disproportionately affecting.

The way they say your name. This changes more than almost anything and is almost never captured on purpose, because it happens fifty times a day and always seems too ordinary to record.

The babble that is not words yet. The pre-language period is completely undocumented in most families, and it is a real stage with its own character. There is nothing to photograph, so nothing gets kept.

The things they believe. Not a sound, exactly, but it only survives in speech. The confident wrong explanation of where rain comes from. The elaborate account of a dream. Written down these read as cute; heard in their own voice, delivered with total conviction, they are something else. Nothing about a photograph carries an opinion.

All of it is gone within months, and none of it is recoverable later. That is the distinction from photographs, where at worst you have fewer than you wanted. Here you have none.

There is also a version of this that applies before language starts at all. The pre-verbal months have no photographable moment — there is no first step, no first tooth, nothing that announces itself — so almost nobody records them. But the babbling stage has its own shape, and a child working through the sounds of a language they cannot yet speak is a stranger and better thing to listen back to than most of what happens afterwards. If your child is very small, this is the easiest recording you will ever make, and the one nobody has.

The habit

Thirty seconds, once a month. That is the whole method, and its only virtue is that it is small enough to survive contact with real life.

The single change that makes the recordings far better: ask the same question every time. Something open, something they can answer at almost any age. “What did you do today” works from the moment they can string an answer together and it keeps working for twenty years. The consistency is what turns a pile of clips into a sequence — the same question, the same child, the answer getting longer and stranger and more articulate. Played in order it is genuinely difficult to sit through, in the good way.

Pick a fixed trigger so you do not rely on remembering. The first Sunday. Bath night. The day the rent goes out. A recurring calendar reminder that just says “voice” is enough, and it is far more reliable than intention.

The practical bits

A handful of things that make the difference between recordings you keep and recordings you skip past.

Record in a quiet room, with the television off. Phone microphones handle voices well and background noise badly, and a recording where you cannot make out the words is not worth having.

Hold the phone lower than you think. Chest height, not up by their face. Toddlers respond to a phone pointed at them by performing, going silent, or grabbing it. Held low and slightly out of the way, they mostly forget it exists, which is the entire objective.

Do not direct them. No “say hello to the camera”, no prompting for a particular word, no second take because the first was not clear enough. The directed version is a performance and it ages badly. The undirected version is a child talking.

Keep the take where they interrupt you. Especially that one. The clip where they wander off mid-sentence, or start telling you about something else entirely, or ask you a question back — that is the one you will play in fifteen years. The clean take is a recording. The interrupted one is a memory.

Say the date out loud at the start if you are recording to a general voice memo app. Files get renamed, moved and stripped of metadata, and four seconds of “twelfth of March, she’s two and a bit” makes the clip self-describing forever.

If you are also keeping a written list of what they say, the two records are better together than apart — the first 50 words and what to expect covers that side of it.

Keep them together, in order

A year of monthly recordings played back in sequence is a completely different object from twelve separate clips sitting in a folder. The sequence is the point — the change is only visible across the gaps, and no single clip contains it.

But that only works if you can find them, and voice memos are exceptionally easy to lose. They are the least-visited corner of every phone. They have no thumbnail, so scrolling past them tells you nothing. They are titled with a timestamp or a room name. Scattered through a camera roll of fourteen thousand photos, they are effectively invisible, and a recording nobody ever plays is not much better than one nobody made.

So whatever you use, the requirement is the same: one place, in date order, labelled well enough that a specific month is findable. A named folder in a cloud drive does the job. So does a note listing what is where. DearBaby’s Voice Year does it by gathering everything you record in a given month into that month on its own, so the year assembles itself as you go and plays back as one piece — but the habit is the article, and the habit works with a plain folder and a bit of discipline.

Thirty seconds a month is six minutes a year. Set the reminder before you close this, because that is the only step most people skip.