Anatopus

Why Your Voice Sounds So Wrong on Recordings

Every line of this video with the sources behind it. I picked this topic, checked the script line by line against the 4 sources below and reviewed every cut; AI helped draft the script and made the visuals and voices.

Watch the video on YouTube

Sources

  1. S1Orepic & Pinheiro (2026). "From Voice to Self: An Integrative Framework on Self-Voice Processing". Perspectives on Psychological Science, 21(5):455–487. DOI · free preprint
    Review

    Used for: Recordings feel estranged and sound unfamiliar, and this hampers research; air and bone conduction and the low-frequency boost, still not fully measured; motor prediction and suppression of the self-voice; pitch perturbation (small shifts corrected, large ones treated as external); the memory template; bone-conducted vibration improving self-voice recognition; self-voice rated more attractive; a calmer self-voice reducing anxiety (small samples); puberty; voice clones

    Cited at 0:00 0:19 0:40 0:50 1:17 1:27 2:11 2:53 3:16 3:36 3:54 4:11 4:48 5:02 5:13 5:26 5:32

  2. S2Schneider & Mooney (2018). "How Movement Modulates Hearing". Annual Review of Neuroscience, 41:553–572. PMC6201761
    Review

    Used for: Middle ear muscles tightening during speech; auditory cortex neurons suppressed during speech relative to playback; suppression beginning before speech onset; the two jobs (hearing the world, catching errors); pitch shifts weakening and, when large, nearly abolishing suppression; stronger suppression for utterances near your average

    Cited at 0:00 1:52 2:11 2:23 2:39 2:53 3:26

  3. S3Blakemore, Wolpert & Frith (2000). "Why can't you tickle yourself?" NeuroReport, 11(11):R11–R16. PubMed 10943682
    Review

    Used for: The link to tickling: the efference copy and prediction that turn down self-made sensations (Video 1's S1)

    Cited at 2:11 2:39

  4. S4Rosi, Lavan & McGettigan (2026). "Self-recognition shapes evaluations of (self-) voice attractiveness". Cognition, 272:106501. DOI · free preprint
    Primary study, supporting S1

    Used for: The 2026 study's design and its split result: recognised own voice rated above strangers', unrecognised rated below

    Cited at 4:26 4:40 4:48

Script

0:00Why does your voice sound wrong?

0:00

Why does your voice sound so wrong on a recording? The microphone isn't the problem. The voice you hear while you talk is one nobody else ever hears: part of it travels through your skull, and your own brain turns it down.S1S2

0:19Why this is weird

0:19

You've heard your own voice more than any other sound in your life. And still, a recording of it can sound like a stranger doing a bad impression of you. Thinner. Not quite right.S1

0:32

The usual suspect is the equipment.

0:35

“That's not what I sound like. It's this cheap microphone.”The usual excuse

0:40

But a phone's microphone picks up the same vibrations in the air that reach everyone else's ears. To the people around you, that recording is pretty much what you sound like.S1

0:50

Psychologists have known about this for decades: people meeting their recorded voice often feel surprised, even estranged from it. It's such a reliable effect that it has made the self-voice hard to study, because the recording scientists play back isn't the voice people know.S1

1:08

So what's missing from the recording? Three things, it turns out. One in your bones, and two in your brain.

1:17How it works

1:17

When you speak, your vocal folds vibrate. That sound leaves your mouth and travels through the air to your ears, the same way it reaches everyone else.S1

1:27

But the vibration doesn't wait for the air. It also travels straight through the bones, tissues and fluids of your head, to your inner ear. This is called bone conduction. It adds low, rich tones to the voice you hear, so in your head you sound fuller. The recording only ever gets the air version. Exactly how much bone conduction adds is still being measured.S1

1:52

And there's a second, smaller filter. When you speak, tiny muscles inside your middle ear tighten, damping the sound of your own voice before it reaches your inner ear.S2

2:04

The second piece is in the brain, and if you watched our video on tickling, you already know the trick.

2:11

When your brain sends commands to the muscles of your voice, it keeps a copy, and uses it to predict what you're about to hear. Then it turns down the part it predicted.S1S2S3

2:23

Recordings from inside the human brain show it. When people speak, many cells in the brain's hearing area respond less than when the same words are played back to them. And the muting starts a fraction of a second before the first sound comes out.S2

2:39

That's the same move that stops you tickling yourself: your brain mutes what it saw coming. For your voice, it does two jobs. You can still hear the world while you're talking. And you can catch your own mistakes.S2S3

2:53

Scientists tested this with headphones that shift the pitch of your voice as you speak. A small shift, and the muting weakens: the brain notices the mismatch, and your voice automatically shifts back the other way to correct it. A large enough shift, and the muting almost disappears. Your brain starts treating the voice in your ears as someone else's.S1S2

3:16

The third piece is memory. Your brain keeps a template of how you sound, built from a lifetime of hearing yourself through air and bone at once.S1

3:26

It's a strict template. The muting is strongest when you say a word the way you usually say it, close to your average, and weaker when you drift from it.S2

3:36

Now play a recording. It has no bone-conducted sound and no buzz in your skull. And there's no movement to predict: you're just sitting there, listening. The recording only partly matches the template, so you recognise it as you, and it still sounds wrong.S1

3:54

Researchers found one way to bring it closer. Play people their recorded voice through a device that also buzzes the bones of the head, and they get better at telling it from other people's voices, probably because of the vibration they feel, more than the sound.S1

4:10The twist: kinder to your own voice

4:11

Here's the part that doesn't fit the story. If recordings sound so wrong, you'd expect people to rate their own voice badly. Some studies found the reverse: people tend to rate their own voice as more attractive than other people's.S1

4:26

A 2026 study looked closer. Volunteers recorded a few words. In a later session, they rated a long list of recorded voices, with their own hidden among strangers', and weren't told it was there.S4

4:40

Those who recognised their own voice rated it above the strangers'. Those who didn't recognise it rated it below.S4

4:48

So the cringe may not be the whole story. Knowing a voice is yours seems to make you kinder to it. Why that is, and why older studies disagree, is still an open question.S1S4

5:01Why it matters

5:02

Your voice isn't just a sound you make. It's part of how you know yourself, so your brain keeps it on a short leash: predicting it, muting it, checking it against the template.S1

5:13

That's why changing it can change you, at least a little. In small studies, people who heard a calmer-sounding version of their own voice during a tense exchange felt less anxious.S1

5:26

It's why the brain has to rebuild the template at puberty, when the voice box grows and the voice drops.S1

5:32

And it's a new puzzle for the age of voice cloning. An AI copy of your voice can sound like you to everyone else, with no skull, no muscles and no prediction behind it. Our Narrator, for the record, is an AI voice. She's never heard herself the way you do.S1

5:50Outro

5:51

The voice on that recording is the one everyone else knows. The one in your head belongs to you alone: carried through your bones, turned down by your brain, and checked against a memory only you keep.

Footage and images