Genso Suikoden Translation help?
- Thread starter SegaXfans
- Start date
The general process (maybe not always) is:
1.) Determine if the text is in VDP1 or VDP2 Ram. (For yabause use the viewer, for mednafen dump the ram and view in something like crystal tile 2).
2.) Determine a write trigger address for use in an emulator.
For vdp1: dump the command list and determine where the command to draw the text is located.
For vdp2: You probably want the address of the pattern table location that corresponds to the ‘O’ in OK
3.) Set up the appropriate write trigger just before your menu is displayed. If you are lucky it wont be periodically triggering due to other events going on in the preceding screen. Then reenter your menu, and the trigger should activate.
4.) The sh2 function you are in should probably be either at the same level or one or two down from the one responsible for printing the text. Start reverse engineering dynamically here. Alternatively save the state and load it up in ghidra and try there.
5.) Eventually youll find something that gets passed in the address of your text and a length to print. Adjust the length appropriately. Youll need to modify the asm instruction or a constant thats buried somewhere. Also make sure that increasing the source length of your text does not have unintended consequences (pointer tables, overrunning ram etc)
1.) Determine if the text is in VDP1 or VDP2 Ram. (For yabause use the viewer, for mednafen dump the ram and view in something like crystal tile 2).
2.) Determine a write trigger address for use in an emulator.
For vdp1: dump the command list and determine where the command to draw the text is located.
For vdp2: You probably want the address of the pattern table location that corresponds to the ‘O’ in OK
3.) Set up the appropriate write trigger just before your menu is displayed. If you are lucky it wont be periodically triggering due to other events going on in the preceding screen. Then reenter your menu, and the trigger should activate.
4.) The sh2 function you are in should probably be either at the same level or one or two down from the one responsible for printing the text. Start reverse engineering dynamically here. Alternatively save the state and load it up in ghidra and try there.
5.) Eventually youll find something that gets passed in the address of your text and a length to print. Adjust the length appropriately. Youll need to modify the asm instruction or a constant thats buried somewhere. Also make sure that increasing the source length of your text does not have unintended consequences (pointer tables, overrunning ram etc)
Malenko
Established Member
All the menu prompts and character names seem to be the "0" file ; for example Barbados is at 0x5FAC8
There are at least 2 glyph indexes , I've attached the index I worked out and if you add 7 to the offset of it you'll get the glyph index values for the 0 file.
It also has my rough ideas of what the control codes are. Ironically(?) I cant find the menus translated by @SegaXfans above. Probably need to search with the +7 offset values in the 0 file, or some other offset I haven't figured out yet
There are at least 2 glyph indexes , I've attached the index I worked out and if you add 7 to the offset of it you'll get the glyph index values for the 0 file.
It also has my rough ideas of what the control codes are. Ironically(?) I cant find the menus translated by @SegaXfans above. Probably need to search with the +7 offset values in the 0 file, or some other offset I haven't figured out yet
Last edited:
Hi,
Small contribution. I explored this game a few years ago (2014) and started a (now dead) thread on romhacking.
If this can help for correction/comparison of malenko's table, here is the character tables I deduced in the past.
It may be incomplete.
That being said, if I recall well, I stopped there because I could not figure out where this damn text was hiding in the files...
It is probably compressed, but I may be wrong. It appears uncompressed in RAM, but similar byte patterns were not in datafiles.
I tried to get a localization in the files by tracing instructions... but failed, I probably lack some knowledge to do this properly (and tools evolved since 2014).
Good luck, and don't hesite to call me for help if you pass this threshold, I am a fan of this game series.
Small contribution. I explored this game a few years ago (2014) and started a (now dead) thread on romhacking.
If this can help for correction/comparison of malenko's table, here is the character tables I deduced in the past.
It may be incomplete.
That being said, if I recall well, I stopped there because I could not figure out where this damn text was hiding in the files...
It is probably compressed, but I may be wrong. It appears uncompressed in RAM, but similar byte patterns were not in datafiles.
I tried to get a localization in the files by tracing instructions... but failed, I probably lack some knowledge to do this properly (and tools evolved since 2014).
Good luck, and don't hesite to call me for help if you pass this threshold, I am a fan of this game series.
Last edited:
Malenko
Established Member
Post the link to the HRDN post? I don't see an attachment on your post.
The first line of dialog in the game (when you talk to Theo) is at 0x16EDA0 in MURA_A1.bin ,
the string is:
DF 81 48 51 55 E8 10 05 00 0E 4C 6D 56 44 48 51 58 47 70 5E 4B E8 04 08 0E 01 E4 01
(I know this thanks to @Knight0fDragon ). This is the line I changed to 012345 as a bit of a proof of concept.
The first line of dialog in the game (when you talk to Theo) is at 0x16EDA0 in MURA_A1.bin ,
the string is:
DF 81 48 51 55 E8 10 05 00 0E 4C 6D 56 44 48 51 58 47 70 5E 4B E8 04 08 0E 01 E4 01
(I know this thanks to @Knight0fDragon ). This is the line I changed to 012345 as a bit of a proof of concept.
Malenko
Established Member
I went to the first line in dialog and ran through all the values, so its likely its just using a different table. At a glace it looks like our glyphs are in the same order for the ones that are there. I might look at the name entry screen to square this away a little more for whomever decides to translate this game.
Malenko
Established Member
Another shout out to @Knight0fDragon ; he showed me that some files are "uncompressed" and have padded 2 byte values, like the start game screen. The menus are in NAME.BIN
Someone should reaaaalllly pick up this game (not it!)
Someone should reaaaalllly pick up this game (not it!)
vbt
Staff member
nice progress !
Knight0fDragon
Patron Supporter
Ideally, we would want to default to English, just need to find the variable that determines the state for the characters.
Thanks Malenko for your work on this. I already know I have bit off more than I can chew but I have now successfully translated all characters with a picture. Basically everyone found in the 0 file.
My next step is to translate them on the stone tablet.
Happy to share the updated files, with anyone wanting to collaborate on this project.
Looking for some help in identifying pointers in other areas. I to have translated the NEW GAME and CONTINUE options but I have been unable to replicate your YES and NO. I can't find the pointer for these. Probably because I'm looking for something similar to the table found in the 0 file.
Any help or guidance on this please?
N.B attached a little video of the early work done. Think I'd only translated 7/8 characters by that point.
My next step is to translate them on the stone tablet.
Happy to share the updated files, with anyone wanting to collaborate on this project.
Looking for some help in identifying pointers in other areas. I to have translated the NEW GAME and CONTINUE options but I have been unable to replicate your YES and NO. I can't find the pointer for these. Probably because I'm looking for something similar to the table found in the 0 file.
Any help or guidance on this please?
N.B attached a little video of the early work done. Think I'd only translated 7/8 characters by that point.
Excuse my absolute ignorance..... what do you mean by set breaks in the debugger? I was able to find the yes/no area in the name.bin (finally got byte search to run) but not the table to change them around to allow for the word yes as opposed to using OK since it is only two letters.
Again I appreciate the help.
Ended up following your Ghidra tutorial.......... absolute game changer!!
Malenko
Established Member
Someone else was asking about this game and I checked my notes and you were right, your values were correct, I fat fingered my sheet.
I also made an "ASCII compliant" font sheet (JIJI.BIN) and added lower case, attached. it does not like the ! and displays it as a (
EDIT: attached the updated glyph table
JIJI.BIN now looks like this:
Last edited:
Knight0fDragon
Patron Supporter
so the routine is a bit weird, but VWF is doable
060230d6 is where the width is pulled in. this grabs a word from 002d682c that is the size. Then there is a check at 060230de to see if the size is 12, and if it is, do the correct routine
This might be enought to BSR to a new area to properly return the width of a given font
looks like this is a global width for the text. This still needs to be traced to see where it actually applies.
060232e2 reads a long from 060ffab8 into r13 and adds it to R1.This is probably where you want to do the VFW, with R13 being width << 16
r11 is the position of the character, r7 is the line numberr8 is the character address that will be read (read as word) so 060232e2 needs to be replaced, basically a new sub that you pass in R8, read as word, figure out the width of that read, perform read <<= 16, return as r13
06023376 is a free word;
at 0602324c, do a mov #e8,r2 and change 0602324e to be extu.b r2,r2 instead. This frees up 0602336e, which we can fill with 1b9c. We have 06023334 point to do a word read at 0602336e, and we now have a long freed up at 06023374 to slap an address into
060230d6 is where the width is pulled in. this grabs a word from 002d682c that is the size. Then there is a check at 060230de to see if the size is 12, and if it is, do the correct routine
what you can do is patch at 060230e4 to change r1 to the desired width
you can change 060230da to be extu.w r1,r0, giving you 060230dc to be a free word (which I change to e108).looks like this is a global width for the text. This still needs to be traced to see where it actually applies.
060232e2 reads a long from 060ffab8 into r13 and adds it to R1.This is probably where you want to do the VFW, with R13 being width << 16
r11 is the position of the character, r7 is the line numberr8 is the character address that will be read (read as word) so 060232e2 needs to be replaced, basically a new sub that you pass in R8, read as word, figure out the width of that read, perform read <<= 16, return as r13
06023376 is a free word;
at 0602324c, do a mov #e8,r2 and change 0602324e to be extu.b r2,r2 instead. This frees up 0602336e, which we can fill with 1b9c. We have 06023334 point to do a word read at 0602336e, and we now have a long freed up at 06023374 to slap an address into
Last edited:
