I don't know how the voices work.
I don't have Mac to figure out.
But I will tell you as a comparison which files the voices we can't use are structured.
Please Keep in mind that this is from an installed NVDA addon, so only parts of the file structure is present.
Usually a Vocalizer Distribution pre v2.x has three Folders: Common, dictionaries and languages.
In Common you have one subfolder named speech with another Folder named components.
On Windows the following files are present in any known Windows Version (note that not all files are present in every build):
clm.dat
lid.dat
synth_med_fxd_bet3f22.dat
sysdct.dat
ve.dll
ve_cdfinit.dll
Apart from These DLL files, all .dat files are universally common, meaning on IOS, Android, Windows and probably on Mac.
These are not the voices.
In the Folder languages you have subfolders named after "language codes", forexample: ged for German, eng for British English or ena for Australian English.
Now I will describe to you the file structure of the British Siri voice named Arthur, which I could not get to work properly.
In the Folder eng is a Folder named speech and in it are two Folders: components and ve
Components contains:
apdb_tp_yosemite_arthur_siri_set1.dat
clc_eng_arthur_cfg3.dat: this seems to be another language configuration file for the voice.
clc_eng_cfg3.dat: this is the main language configuration file; usually there is only one per language; however there are different ones available for each language
rlst_yosemite_set_compact_eng.dat
rlst_yosemite_set_eng.dat
rlst_yosemite_set1_eng_arthur.dat
rlst_yosemite_set1_eng_nav.dat
select_arthur_bet3f22.dat: this is a compact variant voice file
synth_arthur_full_155mrf22_270_06.dat: this is the main voice file for the premium quality voice; this is not the highest quality, there is still premium high
synth_arthur_full_155mrf22_270_06_pmk.dat: I don't know what this file does exactly, but it is required and must be part of the main voice
udct_yosemite_setcfg3_eng_arthur.dat
uselect_arthur_full.dat
userdct_eng.dat
The folder ve contains in this case only two files:
ve_pipeline_eng_arthur_bet3f22.hdr
ve_pipeline_eng_arthur_full_155mrf22_270_06.hdr
These are XML files and can be read and are some kind of basic voice configuration files in which files, and parameters like name, language, gender etc are stored.
These files are not packed or encrypted.
I don't know if all vocalizer voices are packed into one file on OSX.
That's why I would like to know if the System files I described first are present and if someone who has Mac could copy These files and upload them to be looked at.
Obviously we don't Need 700 MB voices, some in compact form would be enough to compare.
Is it also possible to get older OSX builds and maybe run them in a vm to compare file structures, because Images from virtual drives could be extracted.