christos
Oh, the FUN parts.
I initially tried to incorporate past attempts to generate a sudoku 9x9 matrix, even added a backtracking routine that would detect and exclude duplicate solutions. Unfortunately, the whole process of generating a new puzzle was taking about 30". What is more, the outcome, although symmetrical and all, was not what I wanted. The routines were in c
. You can have a look if you like:
Somehow things seem similar (perhaps better) in Greece:
I totally undersand what you are saying. Initially, the original project used local text-to-speech, but was less than perfect, slower and cpu-costly.
You can check it out here https://gitlab.com/christosangel/sapo
Once a FOSS solution gets better and more usable, swapping the tts conversion is not a great deal.