thoughts.txt

# Rehearsal File Generator

Plan of attack: 

1. manually: create MuseScore SATB+ open score files
2. programatically convert to MIDI (MuseScore can do this)
3. programatically generate N (usually 4) variants of the MIDI file with the same settings that I would have in the MSC mixer
4. optionally, programatically perform MP3 render

So we're concerned with steps 2-4 here. 

(2) is easily done, just find all the `.mscz`s matching some criteria and `subprocess mscore -o $outfile $infile` to convert
(3) is doable in several ways, the easiest is probably going to be to use `midicsv` -> for each track in the file do XYZ -> `csvmidi` -> export -> cleanup
(4) should be pretty straightforward using `timidity` or similar (will need a two step process: `timidity`: `mid` -> `wav` and then `lame`: `wav` -> `mp3`)


# Requirements

- python3.5+
- midicsv
- timidity
- fluid-soundfont-gm fluid-soundfont-gs pmidi

# CSV modifications 

- set all track mixer volumes to half initially: `_, _, Control_c, 0, 7, 64`

- Then we want to for each SATB set the instrument as a solo string instrument and make it loud
- if a track isn't assigned to `choir aah` (`_, _, Program_c, 52`) then we don't change it at all
- we'll need to make some assumptions about what tracks correspond to what voice part 
- we can be intelligent about this
- iterate over each track and look at the gamut
- actually, average out the midi numbers of the notes
- if average is above middle C (`60`, use a violin (`Program_c 41`)
- if between G3 (`54`) and middle C, use a viola (`Program_c 42`)
- if below G3, use a cello (`Program_c 43`)
- all of these should be settable
- for each track, set volume to max (`_, _, Control_c, 0, 7, 127`) and set instrument to whatever it should be and export
- oh bother, how do we know which is which - this info isn't in there
- because descants
- for carolling at least the order is DSATB
- or we could just guess based on how many tracks there are
- config with labels? 
- config with labels. 

# Config file structure

- three parts. 
- one for global settings, 
- one for defining patch find-replace (parts)
- one for defining potential orchestration structures (layouts)


Program_c,
# Program operations 

- First do the MSCZ->MIDI conversions

- Ensure the config file makes sense, i.e. layout dependencies
	- thank goodness those are just a DAG

- Then *for each MIDI:*
	- load it
	- figure out if we have a layout match
	- if not, prompt for a layout
	- ensure that all parts are defined as part of prompting for layout
	- *for each track* 
	- working in a track-specific copy of the MIDI...
	- set self's volume to ForegroundVolume, set others' volumes to BackgroundVolume
	- find/replace on self's instruments
	- write out as `filename_Partname_PartNumber`
	- include a comment so we know to ignore it next time!

- Then for each MIDI, convert to MP3

The format conversions should be separate scripts. This is what `find` is for.