Songs that sit inside your vocal range
Give it the lowest and highest note you can sing comfortably, and it shows which songs stay in that band. Not the notes you can just about scrape once on a good morning, the ones you could hold through a whole chorus.
Allow a few semitones of transposition and the list gets a lot longer. Each row then tells you which way to move the song and by how much, which is what you would be doing anyway with a capo or a backing track pitched down.
If you do not know your range, the range test takes about a minute and hands you two note names, straight into these boxes.
Or start from a voice type instead, bass, baritone, tenor, alto, mezzo-soprano, soprano.
If you need something a group can sing together, that is a different and much harsher question, and it has its own page. A room with every voice type in it shares about four semitones in unison.
What the numbers mean
The last column is the smallest move that fits the song into your band. A song wider than your own range never appears whatever you allow, because transposing moves a melody without narrowing it.
The two notes on each row are where the melody spends its time, measured on the sung line and weighted by how long each note is held. A quaver passed through on the way up costs you nothing, a held note costs you everything, so counting notes would have been the wrong measure. The extremes are in the tooltip on each row, and they are usually wider by a third or so.
Where this comes from, and what is wrong with it
The ranges are read out of MIDI transcriptions from the Lakh MIDI Dataset, released under CC BY 4.0 by Colin Raffel. That means human beings typed these melodies in by ear, and they did not all agree with each other.
Of the undefined songs transcribed more than once, 145 have transcriptions that contradict each other by an octave, give or take a couple of semitones. That is 20.4 percent, and the spike sits at exactly twelve semitones, 82 songs against 5 at six, so it is an octave error rather than noise.
Worse, two transcriptions agreeing does not prove the octave is right. 24 songs where the transcribers agreed put the sung line above C6, where nobody sings. Take On Me came out topping at E6 when Morten Harket sings E5. People who type melodies into a sequencer share the same habit of writing them where the keyboard is comfortable, so two of them are wrong together.
So this list only keeps songs that survive two separate checks. At least two transcriptions that agree within a major third, and a range that a human voice can actually produce. 3750 songs went in, 324 came out. 3040 were dropped for having only one transcription, 239 because their transcriptions disagreed, and 39 because the result was outside any human voice. Nothing is silently shifted an octave to make it fit. A correction invented here would be invisible to you, so there are none.
Rows in the wrong octave are the main defect, and here is what I know about them. I pulled 25 rows at random and checked them against songs I know, before any of the filtering below. Five were plainly an octave too high, seven more looked too high, ten were fine.
Then I did something better than trusting my memory of 25 songs. I labelled all 214 acts in the catalogue by whether the lead singer is a man or a woman, which is a fact I am far more sure of than any range. Male-led songs came out sitting higher than female-led ones. In real music men sit somewhat lower, so the male entries were the broken ones, and a third of them were centred above B4 where no male pop singer spends a song.
Those 108 rows are gone. Removed, not shifted. Moving a melody down an octave on a hunch would be invisible to you, whereas a row that is not here cannot mislead anyone. That is what took this list from 432 songs to 324.
It is still not clean, and I would rather say so. After that removal the male-led songs still sit -0.3 semitones from the female-led ones, where a few semitones below would be right. So the male half of this list still runs high on average, and I have no honest rule that fixes the rest. If a song by a male-fronted act looks high to you here, trust yourself over the table.
Treat a row as a decent guess, not a fact. If a song on this list feels wrong in your voice, it probably is, and I would rather hear about it than not.