Good news: I found the type-face (one small pac1 and one big pac 2).
View attachment 332021
Bad news:
View attachment 332023
Now let me explain. Looking at the image above (OG), we see something very strange about the font. So I tried rotating it and we got the next image. As far as I understand, the developer of this game designed the font set in the traditional Japanese writing style (up to down, right to left). This problem is related to editing the font set but is not very important.
The problem lies in the fact that the font set is monospaced (10x10) with non-standard tilesets. As mentioned, the rotation is just for visibility. Both large and small sizes share the same problem. And as far as I have observed, in the dialog box, use a larger font set (pac2) (2 bytes or 4 bytes). So even if I can edit the font set as I want, the display still goes wrong.
Another problem is that the scripts in the game all encode shift-jis. When replacing Japanese characters with Latin, we still have to perform byte calculations on the table and dump the text. And of course, we still use 2 bytes or 4 bytes to change the content, so it is easy to change the pointers, causing the game to crash.
So now what we need most is someone who understands assembly (fast and stable way).
Edited 2: SFF rom construction. (lack details, I will update)
Edited 3: mother3.fobby.net/blog/2008/03/29/welcome-to-text-welding-102/ or we can try pseudo VWF, very classic method to perfect the font set (slowly but surely working).
Edited 4: I figured out how to pac package the data.. However, due to a lack of knowledge about ARM9, there are 2 positions 0x2 to 0x7 in the header I cannot understand. But almost every PAC file I check has the same value.
View attachment 332169