Bach Doodle Dataset

A burst of compositions

Users spent a total of 78.8 years composing the 21.6 million sequences in the dataset, and we heard a lot of popular favourites! The doodle's presets ("Twinkle Twinkle Little Star", "Mary Had a Little Lamb" and Bach's "Tocatta and Fugue in D minor") were by far the most repeated sequences, but we found some surprising runner ups. Explore them below!

The shortest melodies

Out of the top 2000 most repeated melodies, 1347 of them were actually just 4 notes long! Below is a chart of these melodies colored by the the interval after the first note. Warm colours mean that the melody went up; cooler colours means that it was descending. The more opaque the square, the more popular the melody. Hover over any of them to find out more!

The longest melodies

From the remaining melodies, 447 of them were at least 10 notes long:

The hits

We've been labelling the melodies that we recognize! We're hoping some of you will help us find other hits, and this list will grow. Hover over any of them to see what they are.

Once more, with feeling!

A cool thing we noticed is that the dataset contains several variations of popular classics, like "Ode to Joy" (like this full version, with a typo, a different typo, short and slow ). Different people entered fewer or more notes from the same recognizable melody, or got some of the notes wrong. In the visualization below, we are looking at all the melodies longer than 4 notes, and exploring their internal structure. You can also listen to all the prefixes of each melody, even if they weren't actually present in the dataset!



Each cell is a note in one of the top 2000 melodies. Hover over them.