Songs are Easy to Make with Enough Enthusiam & Adrenaline

Andrea and I made a song called "Evil Princess March". Several people have asked about how we made it, so I thought I’d write briefly about its creation process.

First of all, we did this all in one weekend. The total elapsed time from our original inception to final mastering and publication of the song was less than 48 hours. This wasn’t a specific goal or anything, but it was kind of cool to "whip something out" that I think sounds pretty professional.

The actual invested time wasn’t too bad either. My guestimate of the break down is that Andrea spent about 4 hours on it (because she is super brilliant and incredibly talented so she didn’t need a lot of time give a wonderful baseline to the song, without which I could have never even got started), and I spent about 16 hours on it (because I am nitpicky, and, hey, give me some credit, I did all the percussion, sound effects, choruses, theme variations, synth setup, sound design, mixing & mastering). Of that time, I probably wasted 4 or 5 hours just getting back up to speed on advances in sound software since I last did any digital composition. It’s kind of amazing to think that if we were real pros, making a song like this from silence to done probably could be done all in a days work.

Not All Princesses are Evil (Only the Very Best Ones)

At the time of this writing, Andrea is working on a new visual novel game called "Wanted: Dragon to aid in Regaining Throne; No Experience Necessary" in collaboration with some other great folks. We had been talking a little about the main character, Chrysandra, who is kind of an "evil princess". We got to talking about how, in theory, she should have a theme song, and that turned into a discussion about how, perhaps if we weren’t so busy, we’d love to create such a song, and how it was too bad we probably wouldn’t really have time to make it, and somehow an hour or two later, the kids were in bed and we were working on it.

Besides Andrea, some of the best evil princesses I know are from the Disgaea series, and I really thought some of the music from that series had a good feel to it: a little bit anime feel, somewhat complex, but a little quirky and silly. So our first part of "working on it" was listening to some good old-fashioned Disgaea music and trying to saturate our brains with some good ideas.

Getting the Song Skeleton (Out of the Closet?)

Anyway, to start out the song, Andrea wrote up two different single instrument melody parts as well as a chord progression to go with one of them. She did these in Rosegarden, so I decided to just go for it using the same program (even though I was originally going to try using MusE, and eventually planned to drive everything from Ardour). The way these went together, that really gave us two main melody sections. Anyway, I thought we needed at least three, so I wrote a baseline for a third part, in a kind of campy "march" style. I wasn’t sure what melody would go good with it, and I was ready to settle in for a few hours of composing. Right about Andrea asked me to print out the baseline, and then ran over to the piano. With one hand on the keyboard and her other marking music on a blank staff with a pencil, she composed what I consider the song’s signature, characteristic melody in about 25 seconds. I then went back to the computer and entered in into Rosegarden.

As an aside here, if you haven’t ever used a nice DAW like Rosegarden, one if its great features it that you can view any track simultaneously in multiple modes, including classic musical notation, as well as a pianoroll-style matrix. This was great for working together, since Andrea preferred to see musical notation, but I work best with a uniform pitch/duration matrix.

The next part was breaking the song up into sequences. So far we had three melodies, but no real final how the song was going to be structured overall or how we would bridge them exactly. I spent a while playing around, and since our "march" section really had the most "evil princess" character to it, I made that the primary section, and used the other two melodies as bridges between these sections.

Next was adding the two different "squriks" which come in off-beat on the 2nd and 4th measure of each section. They are one of the primary elements to help give the song a "please don’t take this too seriously" feel.

One of the sections that was a too little "boring" on its own I turned into the a silly effects section. This is the part where the princess is experimenting with her minions in the dungeon or something. Most of the effects were as defined in the standard MIDI percussion set, but in one of the sections, I added some sampled effects, such as breaking glass, Andrea recorded a great evil princess giggle. (The one included is like take 22 or something like that, plus pitch-shifted up to sound a little more anime/juvenille). Andrea kept me in check where things had gotten just a little too much over the top, and convinced me to remove a few things that just weren’t working well. You know that awesome whistle? Yeah, that definitely works better very sparingly.

Sugar and Spice and Everything Nice

Once that was done, we had a real song skeleton, and things sounded pretty fun, but the song still wasn’t very rich and was a little dry. I first went through each of the sections and added in tons of minor variations to give it a more human feel. Then I added in an overall base beat with a quirky baseline that kind of leaves you hanging in just the right parts, for kind of a "I am tripping over myself" feel. On top of that, I put in a tick-tock because for some reason I just felt like there needed to be a grandfather-clock feel to the whole thing. So, no, it wasn’t just that I forgot to uncheck the metronome button when I mastered the final recording!

After some listening, Andrea and I decided we didn’t have enough midrange. So I went ahead and added the chanting monks (which of course follow the princess around on her royal duties), and some warm-pad to fill in the midrange a bit, and give everything a more epic feel, because our evil princess is like level 597 or something.

At this point, I thought everything was pretty much done, and I went ahead and did some of the final steps before mastering. The first step here was to go through and adjust all the instruments. The synthesizer I used was FluidSynth. Unless I had a good reason to use something else, I stuck with FluidSynth’s built-in instruments, which are pretty dang decent and realistic as long as you don’t push their pitch limits too far. Up to this point, all of our leads had been piano, so I went through and adjusted them so that we had a good mix of grand piano, xylophone, marimbas, and celestia. I kept most of the base and the primary melody on piano because it just sounded so great.

Next was adjusting the panning. Since everything in our song is digitally synthesized, everything would come out mono without some manual adjustment. I digitally constructed the performance scene in my head (monks over here, piano on this raised platform, xylophone section over here, …) and adjusted the individual channel panning to match. I did some test runs and adjusted the equalization so as not to let any instruments overpower others.

There was a lot of minor tweaking of notes at this point. For example, the attack on the synth choir instruments was a little too wussy, so I had to adjust for it by changing all of the tracks to actually start the MIDI notes slightly ahead of time (about 1/32th note) to make things feel on. I could have adjusted the overall track delay, but that would have caused things to decay too early. Fortunately, since I used a lot of linked pattern copies, this was a pretty easy adjustment.

The Final Step, Mastering

Finally, after some last tweaking of some notes here and there and lots and lots of test listens, I was ready to record the final master. I was originally planning to do this with Ardour, but it was just being too flakey on my system. So instead I went back to my old favorite, Audacity. Audacity doesn’t support MIDI transport, so I had to manually start and stop the record and then trim manually. This wasn’t a big deal, but it did mean that there was a manual step in the process that I didn’t want. Oh well.

Finally, I used Audacity to do final normalization and output to FLAC. Then I used a quick script to encode to OGG and MP3 format using oggenc and lame, and manually updated the tag information with EasyTag. I should have made that part a script, but I was done tagging it in like 2 seconds so didn’t bother.

Now, everything seemed about ready. I had an idea to add a few more instruments in some parts, but (partially) I thought it sounded pretty decent as it was, and (mostly) I was lazy and kind of tired of working on it. I’d been working on this for several hours on while Andrea was out of the house. When she came back, I’m like "sweetie, it’s totally finished!", and she’s like "how about adding some more instruments to all these parts that are boring?". Soooo, I went back and added some more things.

The Final, Final Step, Master, Again

Actually once I got started doing it, I realized that yeah, it definitely needed more. There were too many times we’d go back to the same sequence and nothing was really different. I mean, the song sounded good, but it just felt like it repeated too many parts with not enough variations. The subtle variations I’d added were nice, but honestly not immediately noticable unless you’re listening for them.

So I added a bunch of extra things. I used some of the same instruments from the different melodies and included two new harmonies into the march sections. They play solo in one part each, and then together for the last sequence in the song. I went through all of the melody sections and tweaked notes and added harmonies & chords. I added "courtiers" to sing along with the monks, further filling in the mid-to-high range during the middle of the song. I added "suspicious" heart-beat sounds that don’t quite beat uniformly (aren’t you just wondering what’s up with them?!) in some key sections. I added in some squirky-scratches (really low breath synth sounds). I extended the monks into a couple other parts (they were longly only hanging out right in the middle of the song). I added a final reprieve that fades out, but also allows, with a strategic cut, the song to perfectly, seamlessly loop (in case they want to do that in the game).

Then it was just a simple master of re-panning, re-equalizing, re-master, re-encoding, etc. Finally, we agreed the song sounded "done" enough, and I uploaded it to various places (soundcloud, jamendo), and posted about it on Facebook. The End!

See you next song!

Addendum: Software

Here is some additonal information and links that might be helpful. All software we used is completely free and open source.

  • Rosegarden — The sequencer with many additional DAW features used on this project.

  • MusE — Another sequencer compariable to Rosegarden that I want to try, but not used on this project.

  • FluidSynth — A synthesizer based on Soundfonts, used for all synthesis on this project.

  • Audacity — A full-featured digital audio editor, used for sound effects and mastering on this project.

  • Ardour — A complete digital audio workstation (DAW); not used on this project, but it has a lot of features I like.

  • Virtual MIDI Piano Keyboard — A virtual MIDI keyboard; I actually only used this (and several others) for prototyping sounds; all entry I did with the keyboard & mouse directly in Rosegarden.

Addendum: MIDI Equipment

Andrea has a super excellent full-sized MIDI keyboard synthesizer. But it is not anywhere near my computer, nor is it easy to move around. When Andrea uses it, she takes her laptop over, which works well enough. I could have brought it in the computer room, but it sounded like a hassle.

So, for all the entry I did, I either manually entered notes with the keyboard & mouse. Virtual MIDI keyboards are good enough but slow to use. Not glamorous, but it worked.

I got a little annoyed, though, so after this project was finished, I ordered this to use at my desk for next time:

  • Korg nanoKEY2 — A very tiny (half the size of your computer keyboard) MIDI controller; I didn’t use it to make this song, but I would have liked it. I’ll use it next time around.