I let it set for a day, thinking it over, thinking like a coder, letting my rational brain take over and analyze the situation. Maybe I could shave 2-3k off, get Seven some much needed breathing room. Maybe we could simply cut parts. People would understand wouldn't they? Compromises have to be made in this kind of production. I'm sure the tech savvy demoscene audience would get what we had attempted. So I decided to see where I could get doing some smart cuts.
I didn't keep a good change log, I just saved after every session as I went along. And it started to work! With each cut iteration, I removed instrument pairings (saving note data), simplified the volumes/velocties (reducing entropy by getting rid of note-velocity pairs) and even merging instruments (further removing entropy). I also got rid of those extra patterns I didn't need.
By the 4th iteration I was pretty sure I had cut just about everything that was possible and I had cut a whopping 2,513 bytes!
With going over each track and each instrument over and over again to make cuts (and 105 patterns of this), it took about 4 days to get from iteration #1 to iteration #4. Or an average of 628.25 bytes per day. Each day bought back almost 8% of the size of the total production. It was hard, exacting work, but it was getting us in range.
Very important, even before we got to the first final score iteration, Seven started linking the music in with the visuals and asking me to verify if they sounded correct. This was even before he synced anything, so there were even some crazy mismatches of sight and sound. Doing this early and often became very important to getting the exact amount of space needed and verifying that the music had survived the entire tool-chain intact.
It turns out that this was a very good idea indeed.
10. Pots and Pans
13. Cutting Yourself
15. Pressure Cooker