This algorithm produces a score to perform a given melody on a number of Sing-a-ma-jigs. It optimizes for most notes in a row, then smallest set of dolls needed. It doesn't necessarily find the best possible solution because the number of possibilities increases exponentially with every note. Instead, I used a beam search which saves some fixed number of the most promising solutions, which makes it partially resilient to local maxima but won't take up a ludicrous amount of memory either.
python3 driver.pyThe Sing-a-ma-jigs were a line of singing toys. Each model had one public-domain song that it sang one note of each time it was squeezed. A person could theoretically use many Sing-a-ma-jigs to play a different melody of their choosing. Imagine a handbell choir where every time you play a note that bell will play a predictable but different note the next time. This algorithm (assuming I can get it to work) outputs a score to play a song under those constraints, while optimizing to maximize the number of consecutive notes played by one Sing-a-ma-jig (bell) and then minimize the number of Sing-a-ma-jigs needed.
It also tries transposing the song up and down 12 keys to see if it can find a better match. It uses a little randomization to spice things up, so I make no promise of determinism.
Because they're obnoxious. Also because I saw videos like this one and thought to myself "this person had to do so much planning, and keeping track of the notes each doll makes, and I should make it easy for future generations to do this." And because it's an interesting coding exercise and I need to get more comfortable in Python.
Pardon me while I figure out how dependency management works in the Python universe. Here are the things you need to figure out how to install:
- python3-midi
- the setup file for this required a package called "six" that I had to install separately 🤷
- pytube
import singamajigs
# download all the videos (if you want them in the visualization)
# you only need to run this once!
singamajigs.download_all()
# generate a score object from a MIDI file
score = singamajigs.midi_to_score("path/to/file/myIncredibleSong.mid")A big thanks to the YouTube channel SuperSingamajigs for posting clear videos of each Sing-a-ma-jig's song. This project downloads their videos for the visualizer, but the video files themselves are not distributed with this repo.
The Sing-a-ma-jigs™ are the property of Fisher-Price and I've done my best not to infringe on their rights in this project. I believe all of the songs they perform are in the Public Domain (I skipped the ones that aren't) but if I've posted something I shouldn't have, please let me know so I can remove it.