GBAtemp.net - The Independent Video Game Community

The first thing I did to translate a text was opening the binJ file using Notepad++. Since it directly showed japanese characters I thought, I could directly replace them and expected it to work. Even though I was looking at a binary file, which you usually can't just read with a text editor. But when I copied some texts to Google Translate to find out what they mean, I was disappointed. The texts weren't real japanese. So I googled a bit and found this thread where they chatted about binJ files.

They wrote that these files use a custom encoding for japanese characters and characters in general. What you have to understand about binary files is that they are basically 1s and 0s. This is true for all files, but "normal text files" use a standardized encodings. This means that the byte "0110 0001" with a common encoding like ASCII or UTF-8 always stands for the character "A". But for the game I tried to translate a custom encoding is used. They encode the character "A" using the byte "0000 1101". Once I used the encoding table provided in the thread I was able to read the correct texts. Now I tried to change a single character by replacing the original bytes for this character with the bytes for the character I replaced it with. I sent the file to my 3DS and launched the game. It worked. Great.

Now I only needed to find a way to tell the individual texts appart and luckily the texts were always separated by a special two-byte long token. Thus I wrote a small programm that read the binary data, splitted the texts and decoded the bytes to japanese texts. And it worked great. This was the point where I started developing the first version of BinJ Editor. A thing that I haven't talked about yet is that with this method you have to keep the byte length of the original text and the translated text the same. Otherwise the game won't find the texts and just crashes. The problem with this method is that japanese is a lot denser than english. Normally two characters of japanese text are translated with about eight characters of english text. But since the encoding table they used is optimized for japanese text, a lot of common japanese characters are encoded using no more than a single byte. This means the english translation for a two character long japanese text is only allowed to have two bytes as well. As an example you could translate "耐性" (resistance) with "RE" but not "RES" and definitely not with "Resistance". That was a thing I was able to overcome much later.

But what I was able to do was shorten the texts by simply replacing not-needed characters with null bytes. And therefore I started developing the first version of BinJ Editor. After the first sketch was finished I started translating a few menu texts as a proof of concept. And this was the point were I realized all binJ files have a weird long "first text" that I never looked into. All first texts looked similar and at this point I thought maybe those bytes are the addresses of the texts in the binary file. I checked it and I was right. That means that given a list of texts (with their respective lengths) I was able to re-create this address block. Now I had two possibilites: rewrite my import function of binJ files to validate that the texts I obtain by looking at the separator tokens are correct OR just believe it works. And... I chose the second option. Which was the bad one. But at least I could now replace texts with translated texts of arbitrary length. So I rewrote a lot of parts of BinJ Editor and started translating more parts of the game. Until...

... there were problems. I translated some battle messages and whenever I entered a battle the game crashed. So I looked into the binary data once again and found out that this particular binJ file included the same address twice. So I had to look into the address block while parsing binJ files... I started rewriting the corresponding parts of BinJ Editor and at some point I realized that I always missed the first text of the translation before because the first text does not start with the separator token as all the other texts do. And did I mention one file contains a one-byte off pointer? Anyway... someday I got all of this to work and this basically is the whole story of how I got to decipher the binJ file format and how I wrote BinJ Editor. I later cleaned things up, added some convenient features, wrote documentation and... created this post.

Note: The monsternames are fixed, so my Slimey has a japanese name and the others have german nicknames.

Note: The game does weird things with the width of some texts, so "AGIL" displays unreadable and "A G I L" does the job.