@plasturion Thank you again! I'm completely new to rom modding, so your descriptions help a lot.
@DrYoshi I don't know about others, but I was thinking about it and I'm not very convinced, because: 1. finishing my LA3 translation will probably take me at least a year more, considering the amount of free time I have and that I write down and try to learn every new romaji word I encounter. 2. Infinity was not developed by Imageepoch, but by a different studio, so it seems more like a non-canonical spinoff to me and reviews I've heard/read are pretty bad, compared to the rest of the series.
Anyway, for the last couple days I was focusing on ways to put the translated script back into the game and to my surprise I was able to get better results than I initially expected (considering how noobish I am).
I cross-analyzed partially translated rom by Yuukiwa-kun and Luminous Arc 2 and it seems that LA2 uses pretty straightforward method - every latin letter is represented by a standalone address referencing the font file. For example:
88 C0 -> A
88 C1 -> B
88 C2 -> C
It seems pretty logical, but the game mechanic had to be modified for it to work that way and if you use font file from LA2 in LA3, you encounter a problem of a single letter taking double space (which is visible on the picture I posted above) and possibly taking too much space in the rom in the end. That's why LA3 mechanic (and the script layout) would have to be modified in similar way as in LA2, but that's impossible with my "skills".
Yuukiwa-kun in his partially modified rom used another approach, which was redrawing fonts and using character pairs, for example:
88 8C -> Le
88 8D -> vi
88 8E -> Gl
88 8F -> en
This allows to save space, but requires putting all combinations of small/big letters, punctuation marks and maybe more (possibly numbers) in the font file. Since there are 26 letters in the English alphabet, the number of lowercase character pairs would be 26*26= 676.
When you add uppercase, space and ,.?! you end up with 57 signs -> 57*57= 3249 combinations.
It seems to me that Yuukiwa-kun was redrawing those pairs he needed manually without any order. He ended up with around 1000 character pairs in the last publicly available rom, which is insanely many, but still not enough. Maybe that's one of the reasons he eventually gave up with translation.
Anyway, to follow his approach in orderly manner, someone would have to write a script which would generate an image file with all possible two-signs combinations, looking like this:
Then, another program would be required to encode English translated script into addresses of those character pairs and possibly keep the length of script file the same (by adding some white space padding). It would also have to modify the pointers at the beggining of each script file so that they point at the new starts/ends of translated-encoded messages.
It's all doable, but to me seems like a major pain in the ass and I don't know what other issues might arise after achieving it. That's why I will give up on this for now and rather continue with my translation. I can provide anyone who would want to deal with this with all the details I currently know.