In the end, I used another tool to create the .wav files from text. TTSautomate uses text-to-speak, you can type in the text you want end get .wav files.
I'll add a screen print below. I used a "strange order", because the packs are not selected in the order of loading, but in their alfabetical order. By putting these numbers in "phrase to speak" the selection goes 1,2,3,4 etc. I could get my files sorted and have the filenames reflect the soundpack sequence numbers, but in the end it doesn't matter.
But.... in which directories to put them? It is sort of unpredictable. I put the appropriate .wav file in a directory 0000000000-Announce in "music" and "voice". I also add a similar directory to "single", but place an .wav file there that doesn't produce any sound. As always, I remove the buffer file, which is then created by the card. The 000000000 "just" makes sure it is the first in the list.
Whereas for many sound packs this works great, for some it doesn't. Difficult to see the rules, some announcements of the soundpacks keep on playing effects or music. What is the rule here? Anyone knows the magic that is behind it? In the end I prefer a "soft" music file reflecting the theme overlaid by my generated .wav file reading out the soundpack.