I sat down with my friend Richard Gould at Designing Sound to talk about some of the nuances of working on and designing the audio systems for the game Mini Metro, as well as share some early prototype footage, recordings and documents.
I sat down with my friend Richard Gould at Designing Sound to talk about some of the nuances of working on and designing the audio systems for the game Mini Metro, as well as share some early prototype footage, recordings and documents.
I’d like to talk about some of the techniques I explored in creating the soundtrack for Monsters Ate My Birthday Cake. I have used some of these techniques before, but others represent a new approach, and I’m excited to talk about them.
With each new project, I try to reinvent myself in some capacity. This time around, I devised a plan early on that would take me in a new aesthetic direction. Games like Yoshi’s Island and Wind Waker inspired the developers, and I found it fitting to pursue a similar path.
The song ‘Alone in Kyoto’ by Air inspired me to use this technique. It took me years to realize that around 1:20, there is a three-note figure that sounds in three sequential instruments.
In 'Yumberries,’ there is a three and sometimes five-note figure that hockets between two instruments. One is in the left channel and the other in the right. The right-panned instrument has a bit of a pitch drop as it ends, and the other has a plucked, string-like attack. The instrument on the right side always plays the third note in the figure. See if you can hear it! The two timbres are similar, so it’s a subtle effect.
Melding the styles of different genres is part of the charm of games like Super Mario 64, Animal Crossing, and Earthbound. I tried to follow that pedigree while pulling from my external influences. There are backbeat chords (Reggae), claves (Afro-Cuban), and shuffles, to name a few.
I saw this project as an excellent opportunity to tap into my long back catalogue of ideas. I was able to repurpose more than 20 ideas that have collected dust on my hard drive. Some date back as far as 2004, the year I first started writing. I’ve included them with the download of the soundtrack from Bandcamp.
To keep the experience cohesive, I listened to hundreds of sketches, always with the intent to find a particular usage. Sometimes during the process of updating a sketch, I would decide it was no longer appropriate for the placement I had intended. When this shift occurred, I would set it aside, and wait for the proper placement to arise. In most cases, this worked, but there were a few sketches that turned out not to fit.
Delay is a formidable tool for creating new parts. In ‘Land of the Blue Muckwic,’ I shifted a part that was once doubled back a 16th note. I also reharmonized the part to create a distinct counterpart to its original. ‘Sleep Tight, Little Buddies’ uses a similar technique. It takes a single part and turns it into two discrete elements.
I found the notion of a 42-track album of looping loops scary and set out to vary the structure of songs when it was beneficial. Many songs have endings unique to the soundtrack (i.e., ‘Scent of Betrayal,’ ‘Yumberries’). Some have removed or reordered sections on the repeat to mix it up (‘You Got My Rollerskates,’ ‘Ruins of the Firefather’). Others make use of new combinations of parts to strengthen the variety on display (‘Ducky vs. Candy Quarterback,’ ‘Poot of Gricklesmudge’). I took care to go beyond having a soundtrack of simple loops. I am a strong believer that a soundtrack and a score are not the same, and I did my best to treat them as such.
This technique isn’t surprising or new, but I find that a well-placed pitch bend can add interest to a part that might need it. Eirik Suhrke is the composer for the game Spelunky, and we came up together in music. We ran a chiptune netlabel together from 2007 to 2012, and his work has left a lasting impression on my own. His creative use of pitch bending is one such effect that’s rubbed off on me. I can sense myself channeling his style in certain spots on this soundtrack.
'Yumberries’ moves between straight and shuffled beats. The shuffle gets stronger over the course of the song, and then subsides to help create a smooth looping experience.
I discovered a method that allows me to play in segments that have a fluctuating tempo. I can then beat map the remaining song elements to that performance. This song along with a few others on the soundtrack have live performed ritardando endings.
I’m not sure exactly where I’ve heard this technique before, but it produces a pleasing sound. On the final chord of a few songs, the notes die out in a staggered fashion, often from the highest pitch to lowest. For lack of a better description, this sound is something akin to water sucked down a drain.
This trick often pays dividends. I took the upbeat Nogport Meadows music and reharmonized it as an evil sounding piece for The Deadlands. In a tight spot, this can be a justified way of cutting corners.
I set out from the beginning to assemble a large library of sounds that I could reuse throughout the soundtrack. On previous projects, I stuck to an aesthetic by reusing sounds from memory, and discovering new ones. Having a directory of sounds to work from and build up, helped get production out of the way early and often. This approach allowed me to focus more on writing.
I looked to games like Yoshi’s Island to help frame my aesthetic decisions. To capture a 90’s tone, I gravitated towards clean, short sounds. The palette is a blend of synths, SNES soundfonts, drum machines, & beatbox samples. I also used hi-fidelity sample libraries, repurposed to seem simple and cheap. I went into the guts of many of these sounds to shorten their volume envelopes, often to a drastic degree. This method proved to be an effective way to mimic the character of one-off samples often used in SNES and N64 soundtracks.
Traditional reverb was almost never the right choice. Instead, SNES-style echo proved to be an effective way to add depth to some of the pieces. I attained something that approximates this with a short delay time and a high feedback setting. I also took into account the limitations of the SNES sound when useful. The sampling rate of the S-SMP (SNES sound chip) is 32kHz, so at times I would low pass samples down to 16kHz for a muted character.
There are a lot of elements scattered throughout the score that pay direct tribute to Koji Kondo.
'Scent of Betrayal’
This song has an upper structure triad on the last chord, which you can also hear in the game over music from Super Mario World.
'Tolerable Levels of Trespassing’
The occasional flute trills in this song are a knowing nod to the Forest of Illusion map music from Super Mario World. The interior house music from Wind Waker inspired the musical style.
'A Trail of Crumbs’
This track started as an aesthetic study of the Wind Waker theme song and evolved into something else. The opening strums by a zither-like instrument remain.
Yoshi Drums
Koji Kondo makes great use of Latin percussion in many of his scores. The most notable example is the “Yoshi Drums” in Super Mario World, where percussion appears when Mario sits on Yoshi. I tried to capture some of that charm.
'Terror, Esquire’ is one of my favorite tracks on the album. The seed of this song started as a sketch recording of an electric guitar. In most cases, I built on sketch ideas; I reperformed them or imported their MIDI data when appropriate. This track is the exception to the rule; it’s the only song that features a live instrument. I used a few tricks to help glue the parts together. In continuing with the theme of clean staccato sounds, I used a transient shaper to reduce the sustain in the signal. I also beatmapped the song and its elements to the guitar performance, instead of quantizing the guitar. I believe this gives the song a unique groove and allows it to stand out as a highlight of the score.
This project took a circuitous route to completion. I made contact with the developer in January of 2013 and created a few sketches for the Kickstarter in February. We didn’t start writing music for the game in earnest until January of 2014, a full year later.
In an ideal world, I would be able to focus on one project at a time, and embed myself in the development process from the beginning. This would allow the music to evolve with the game, and inform early decision making in other disciplines such as art and design. The unfortunate reality is that the creative and financial economics don’t always add up. Devoting myself to a single project for long periods is often unsustainable. It’s been my experience that I can only give a project my full attention for a few months before I burn out and need a change. As a result, it has benefitted me to jump onto projects later in the process, and to change my focus from time to time.
This project follows that trend. I created most of the material for the game in two spurts. As you might expect, the first spurt happened before the game came out. This spurt covered most of the basic music (map, title, village, levels, etc.). There was ~15 minutes of music in the game at release, and I found this unsatisfactory. In most cases, there was a single 60 to 90 second loop in use for 15 - 20 levels on its own. Also, I never found the time to match the amplitude readings of the various tracks. While most of the early reviews seem unfettered by the quantity of music in the game, there were comments about the volume levels.
Having a year and a half and still being unable to deliver the proper amount of content in time for the game’s release is a result of a few things. As a freelancer, I often sign onto projects well in advance of when I expect to start working. This tactic is in part due to how long some of these projects are in development, and also a scheduling tactic on my part. In recent years, I’ve tried to focus on one project at a time for as long as possible before switching. I’ve found this reduces stress and makes my work more cohesive. I will often wait for the best time to begin, based on my availability, the progress of the team, and whether there are playable builds. There are times when more than one project sets its due date around the same time. When projects coincide, I must figure out how to work on them at the same time while still meeting all their deadlines. This conundrum is why Monsters Ate My Birthday Cake only shipped with 15 minutes of music.
Unsatisfied with the amount of work on display, I went back and played through the game quite a bit. I took notes and tried to figure out all the places that required new music. I spent the next month or so writing, and by the time of the first big update, the soundtrack length had quadrupled.
In an attempt to create a musical balance, I developed a technique to approximate how often each track plays. Without the actual numbers to figure that out, I settled for the number of levels assigned to each track. Here are two pie (or should I say cake?) charts, to show the music usage. These cakes only account for interactive levels, not cutscenes or map screens:
After the update, I was able to reduce the burden on each piece quite a bit. Track 01 and 02 of each zone trade off in code, so that you’ll never hear either twice in a row. The rest of the tracks belong to special levels. Special levels include those that introduce new monsters, side paths, unique encounters, etc.
I applied this same approach to cutscenes as well. On release day, most of the cutscenes used the same music. As of the update, every cutscene has a unique blend of tracks.
I also introduced music for the “Dark Realms”. This gameplay mode occurs when the player enters a ground tear. The screen loses its color, and you have access to a different version of the level. The tracks in these sections are variations of the original music, but with a few differences. They are frequency shifted, pitch modulated, and lower in volume. These changes add to the variety even more.
Another addition was musical progression for the map screens. Each zone has a map screen where you access the levels, and as you unlock new levels, new versions of the music unlock as well. The map music changes altogether when you unlock the next zone. Here is a breakdown:
Map Music, On Release
Five tracks, five assets. One track per zone and one for the world map. No variations.
Map Music, After Update
Nine tracks, twenty-four assets. Two tracks per zone and one for the world map. Four progress-based variations per zone.
Before the game’s release, a black screen intro provided backstory, and I wrote a piece of music for it. I thought it worked quite well, but we removed it to allow the player to jump into the game quicker. Instead of letting the track go, I managed to find an interesting place to use it. There are bottles scattered throughout the game that contain messages. When you discover one of these bottles and read the note inside, the background music dies out and a short piece of music plays in its place. The idea was to characterize Niko (the protagonist) as becoming immersed in the reading of the message. It is a somewhat unorthodox usage of music (reminds me of the game ‘Ōkami’), but I believe it adds a little extra immersion to the game.
I have given some thought to having in-game assets mastered over the last few years. That said, when the time came to make a decision, I decided to forgo the option. I tweaked assets often, and until the last moment. I find the concept and workflow of mastering, at least in the traditional sense, to be somewhat incompatible. I searched for an alternative that would get me at least some of the way there, and I found Leq. Leq is the Sound Pressure Level in decibels, equal to the total sound energy over a given period. While not a silver bullet by any means, Leq allowed me to bring the assets close together in perceived volume, in most cases. I chose to use A-weighting, which uses a response curve that favors higher frequencies, much like our ears. I also found it matched up well with the response of mobile devices, which tend to be poor at reproducing lower frequency sounds. This was a factor because the game is available on iOS and Android. To calculate Leq, I used a tool called AudioLeak, which will output a single dBFS value. I used these values to attenuate the signals of the files to desired levels.
There is one idea that didn’t make the grade. I wanted to have the music change depending on which monsters were at your disposal. In levels, music would crossfade between different versions as you switched between monsters. We also toyed with the idea of changing the map music in the same way, depending on which monsters were available. The levels concept would have worked well, but I decided to dismiss it and focus on quantity instead. I also planned to write a jingle for each monster that would play if you beat a level with them, but never got around to it. Some ideas, while good, are not important enough to use.
All in all, I am pleased that I took the time to beef up the music in this game. I can look back at it and be proud, instead of having reservations and wondering how it could have turned out better. I’ll always have some of those what ifs, but by giving the music the extra love, I am far more content with the result.