Blend
Hear
Play the clip. You hear the melody and the line you are about to sing against it, and you can slow it down, loop a bar, or hear either voice alone.
Sing
Sing your line into the microphone while the melody plays. Your notes appear on the staff as you land them. Nothing is uploaded; the analysis runs in your browser.
See
A score out of a hundred, stars, and a breakdown by kind of miss. Then the same clip again if you want it, or the next one.
Modes
Three ways to practise. Lessons teach; Perform tests; Intervals trains the ear with no singing at all.
Lessons
84 lessons across 23 concepts, in nine stages. Each lesson is a few clips on one idea, read first and then by ear. A checkpoint closes every stage: a handful of clips from everything so far, shuffled, nothing written out.
346 clips, 46 of them songs with their words and the rest practice lines written for the job. Every clip is served at the length the stage is ready for.
Perform
A scored run. Each clip is one take. The difficulty goes up when you sing cleanly, holds when you scrape through, and steps back when you fall apart. Five hearts, then it ends and tells you where you stopped.
A daily run gives everyone the same clips, one attempt, and a leaderboard.
Intervals
No microphone. A tone sounds and holds, you are asked for an interval, a fifth above or a third below, and you turn a wheel until the second tone blends with the first. The wheel has no note names, and the notches only light up once you commit.
A clock runs down and only accuracy buys it back. You are graded against the beatless interval, the one a second singer tunes to, which on thirds and sixths sits about fourteen cents from where a piano would put it.
Why it is built this way
The lessons follow what is known about how people learn to sing in parts. Each of these is a design decision with a reason.
Singing the note instead of naming it
Recalling strengthens memory more than recognising, and producing more again. Ear-training apps ask you to name an interval you heard. Every round here is sung, because a second voice is something you make.
Feedback on every note as you sing it
Practice improves you when the feedback is immediate and specific. Each note colours on the staff as you sing it, and the breakdown afterwards says which were flat, which were late, and where you sang the tune instead.
Help that is free at first and withdrawn later
The staff, the first-note cue, half speed and hearing your line alone are free early in a unit and priced later. The last clips of every lesson, and every checkpoint, are sung by ear with nothing written out. The notation is a scaffold for the ear, not the thing being tested.
The order the lessons come in
Unison and octave, then a drone under a moving tune, then thirds and sixths, then fourths and fifths, then independent lines and canons. Part-singing has been taught on this ladder for a century; the lessons climb it one rung at a time.
Lessons on one idea spread out, not drilled in a row
A unit’s lessons are spread through the course rather than taught back to back, and each checkpoint mixes everything learned so far. Harder in the moment, better retained.
Harmonising with a recording of your own voice
Research on singing accuracy finds the usual fault is between what you hear and what your voice does, and that it improves with feedback-rich imitation. Half the clips have you record the tune and harmonise with your own take, which is that loop at its tightest.
Common questions
I think I might be tone deaf.
Almost nobody is. Of people who test themselves, about two in a hundred cannot tell pitches apart. The rest can hear the difference and have not yet trained the voice to follow the ear, which is what the lessons do.
I cannot read music.
You do not need to. The staff is shown as a picture of where the line goes, up or down, and you can hear the line instead of reading it. The later clips have no staff at all.
Will my laptop microphone work?
Yes. Any built-in mic is enough, as long as you wear headphones so it hears you and not the melody coming out of the speakers.
Where does my voice go?
Nowhere. The pitch analysis runs in your browser and no audio is uploaded. The duet takes are kept in memory for the length of the clip.