Slept badly and awoke late, my mood strange and sad, perhaps due to the end of projects.
I started the day by updating Prometheus again, to v3.23, this time by making my spectral EQ-Tracker use RMS values rather than peak values. Peak values are easier to calculate, my reasoning was that the mix of peaks of different frequencies would give a good measure of the overall feeling and balance of a track. The problem is that one loud transient (a crash or boom) will set those peaks, rendering all of the rest of the song irrelevant. So I updated it to use RMS (average) values.
This wasn't too difficult, the program already calculates them to return the overall level. Doing this on a large sample can lead to inaccuracies due to the huge numbers involved (adding tiny numbers to a floating-point variable 10,000,000 times or more), so I divide the calculation into smaller sub-groups and average those groups at the end, and I use 64-bit double precision.
This meant creating a new 'standard' as an audio target for matching the levels, and thus loading in lots of existing songs and viewing the relative levels. The results were, actually, almost the same as when using peak values, but not quite. There's a definite arch shape: low, high, highest, high, low; whereas the peak traces were more low-flat-flat-flat-low.
Of course, the important thing is how it sounds. I reasoned that a mix of an infinite number of musical tracks would produce pink noise, and that music engineers would prefer white noise. If this hypothesis is correct then all EQ balancing is ultimately a matter of tuning pink noise to white; removing bass and adding higher frequencies. This seems to match my experience.
After that I started work on the Sky Robes of Celeste production. I started with the piano and added a few strings with minimal filtering. The whispery vocals clashed a little with the piano and the somewhat hissy strings, but it sounds good enough. I started the music by painting a scene, starting with a live recording taken from the Locus Sonus world microphone array, this one in Dar es Salaam at 4pm today. This forms the gap between day and night, the living and the dead, of this poetry about ancestors and graves.
