If you have not yet seen the original post about my Merry Go transcription for player piano, please read about it here.
The transcription was initially organized in four voices in the Lilypond source file. This made it relatively easy to direct Lilypond to write out a couple extra scores for a single-piano duet version. The two treble voices can be played by Primo, and the two bass voices by Secondo. This shows the power of Lilypond for maintaining multiple renditions from a single source file.
The duet was still not immediately playable. A few chords had too many notes and needed minor tweaks. The last 3 measures needed an alternate playable version; I think I struck a reasonable compromise by having the players alternate the fast final runs, and by reducing the 3-octave run to 2-octave. And in this version, the final chord does not contain all eight C keys on the piano!
As of this writing, the duet has not been performed by two people. I did record myself playing Secondo and then played Primo over it, but there may remain some awkward overlaps with two players on one piano.
If you’d like to print the 6-page duet in book form, I would recommend printing page 4 as simplex, pages 1 and 5 as duplex, pages 2 and 6 as duplex, and page 3 as simplex. Then you’ll be able to have Secondo on the left facing Primo on the right, for 3 pairs of pages. The page numbering is designed with this scheme in mind.
In 2013, I learned to play the Sousa/Horowitz piano transcription as best as I could. I found three sheet music transcriptions (Pellisorius Editions by Jon Skinner; Christian Jensen, 1999; Florian Wolf, 2010). Each edition differed from each other, and also differed from the available audio recordings in minor but unsatisfying ways.
I took it upon myself to create a new, highly accurate transcription of one particular performance by Horowitz: the seminal April 23, 1951 radio broadcast from Carnegie Hall. I attempted to capture as much detail as possible all aspects including notes, dynamics, pedaling, and tempi. I even tried to capture his apparent mistakes, which I have largely indicated in the music. (Of course, this being his own version, “mistakes” is a very subjective term.)
Once again, I relied on Transcribe! software for listening and GNU LilyPond for typesetting. The process took many days.
If you’ve watched enough compilation videos on YouTube, it is likely you’ve heard Merry Go, a fast-paced, 2-minute ragtime-ish piano song published by Kevin MacLeod in 2010. It is one work out of a large body of high quality digital music that Kevin shares under the Creative Commons By 3.0 license, for anyone to use for any purpose, royalty-free. Not surprisingly, his music has become some of the most-used background filler on YouTube.
Depending on your mood and temperament, Merry Go can be either energizing or irritating. Because of the latter, someone had to do it: here is a video of the song playing for 10 hours. There is also a 1-hour version, a slowed-down version, MIDI transcriptions, and others.
Watching a useless video late one night, this song played, and it piqued my interest. Its use of sforzandos and staccatissimo is striking. It’s fast, but sounds simple enough. I decided to attempt a sheet music transcription. However, I quickly found out that the apparent simplicity belies a greater complexity than one would expect. There are a lot of notes! It got even more interesting as I discovered its origin..
A reproduction aiming for fidelity to the original would be most suitable for performance by MIDI and/or player piano. It is impossible to play on solo piano.
Using the Transcribe! software, I listened to Merry Go for hours in excruciating detail, at different speeds and octave shifts, in order to generate a transcription. To test for the presence of specific notes, I would listen carefully while hitting them on a keyboard while simultaneously looping short snippets of the audio, often just individual chords. There are several things that limit how accurately the notes can be divined. Some chords contain upwards of 10 simultaneous tones. Very quiet or low tones may get drowned-out. Artifacts of digital encoding and frequency scaling through re-sampling filters result in ghostly sonorities and sympathetic octave blurring. It took four or five passes/editions before I was satisfied I hadn’t missed too much. Nevertheless I’m sure Mr. MacLeod would cringe.
Having written out the initial transcription on music staff paper, I then coded it as a text file in the GNU LilyPond music description format. LilyPond is an amazing piece of software that reads a description and generates beautiful typeset sheet music in PDF format, as well as generating MIDI output. There is a big learning curve, however. I’m sure it comes a lot more naturally to people who are already experienced computer programmers. But if you are a geek who needs to typeset music, you need LilyPond!
Here is the resulting MIDI output. Dynamics processing by LilyPond is somewhat rudimentary and leaves something to be desired. As with everything else in LilyPond, perfection is possible given enough time (worst case, it’s open source!)