Hold the Note — a singing trainer that stops until you sing in tune

A note plays, then stops. Sing it back. The melody does not continue until you are inside the tolerance and have stayed there — touching the note is not holding it.

Your microphone feeds a pitch detector running inside this page. It shows the note you are aiming at, its frequency in hertz, and which way you need to move. Nothing is recorded and no audio leaves your browser.

A hundredth of a semitone; 1,200 to an octave. Cents are used instead of hertz because every octave doubles the frequency, so one semitone low in your range is about 13 Hz and the same step an octave up is about 52 Hz. Cents mean the same thing everywhere on the scale.

A4 — the A above middle C — is 440 Hz, and everything else follows from it. Middle C is 261.6 Hz. Each octave doubles: A5 is 880 Hz, A3 is 220 Hz. The target frequency is on screen for whichever note you are singing.

Yes. Any octave counts, so a bass can sing a line written for a soprano and still land it. You are judged on the note, not the octave.

No. The microphone feeds a pitch detector running inside the page. Samples are overwritten by the next frame; nothing is stored, uploaded or sent anywhere.

Not by playing the recording — that is a licensing question rather than a technical one. Everything here is synthesised in your browser from public-domain melodies and exercises.

That is the whole idea. Feedback you can ignore gets ignored, so the melody waits until you are inside the tolerance and have stayed there. Touching the note is not holding it.

It will not play your favourite song. Every note is synthesised here and the melodies are public domain or exercises. Shipping somebody’s master recording is not something an experiment gets to do casually.

It is not a tuner. The detector expects a sung vowel held for most of a second. It will lose you on consonants, on breathy notes and in a loud room, and it says nothing rather than guessing when unsure — a confident wrong note is worse than a blank.