GameMaker hook not working with VA-11 HALL-A

[v2lmmj04]

I'm trying to hook VA-11 HALL-A with the latest Chenx221 release (Textractor_260130.zip), but I'm only getting random characters:

My system locale is Japanese and the language chosen in the game is also Japanese.

It was mentioned in this thread that this specific game was tested for previous alpha builds from 2023:

#868 (comment)

GameMaker engine hook (PR #1085) tested with "VA-11 Hall A"
With GameMaker hook use the Regex Filter extension
.+\]
or Regex Replacer extension
|REGEX|.+\]|BECOMES||MODIFIER||END|

But it seems to no longer be working as expected.

[Chenx221]

@v2lmmj04
The random characters are caused by an encoding mismatch (UTF-8 vs Shift-JIS). Please try these two methods:

• Replace texthook.zip and use GameMaker hook with Regex Filter: .+\]

• Alternative: Use H-code: HSN65001#-4@D4FFF:VA-11 Hall A.exe (with Regex Filter .+\]|#)

Note: Remove any redundant filters to avoid interference.

---

Additionally, the previous hook might miss some dialogue text.
There is another method: you'll need to replace the game's data.win file with this, and then the text will be sent to your clipboard.

[v2lmmj04]

@Chenx221 Thanks for such a quick response!

I thought it might be an encoding issue, but when I noticed some lines weren't even generating any text in Textractor I got confused. But I guess based on what you said it turns out it was two issues together then

I tried the HCode method and removed all of my other extensions (they were causing some text to be lost as you probably foresaw), which got it no longer being random characters but yeah it was still missing some dialogue completely as you mentioned (e.g. the first line of the game "アナ：ねえねえ。", the "アナ：でも人生の９０パーセントは受け止め方しだい！物事の明るい面だけを見るようにすれば大丈夫。" line shortly after, etc.)

I tried the last method you recommended, replacing the data.win file of the game, and that seems to work best. At least in the prologue, all of the text got sent to the clipboard properly:

Opening the modified file in UndertaleModTool, it seems the main change was just around the gml_Script_textbox_cmd_load script to add a clipboard_set_text(string_replace_all(temp_text, "#", " ")); line before returning?

I'm a total noob at reverse engineering, so I'm curious -- is it just not possible to externally hook that same function to pull the text out (then do a similar text replacements on it) with a tool like Textractor (or a similar tool that may allow more custom processing like agent)? For why the original hook misses some dialogue that is

Even though it's not able to send over websockets this way, it should work for me now though. Thanks so much!

[Chenx221]

For why the original hook misses some dialogue that is...

The original hook was actually attached to the GameMaker engine's string_pos function.

Because the game script is designed to pre-load dialogue, string_pos triggers frequently within a short window, carrying large volumes of text. When the accumulated text length exceeds the buffer limit set in Textractor, the string is ignored.

Even though it's not able to send over websockets this way...

You can switch the input source to 'Clipboard' in Textractor; it should still pass the text to your websocket plugin, provided the plugin isn't configured to ignore clipboard data."

---

Modifying the data.win file isn't 100% effective. If the game was compiled using YYC, the game scripts are converted into machine code, making them impossible to modify easily.

[v2lmmj04]

Oh, I see. So in theory I could increase my buffer setting, or I could find a hook for the gml_Script_textbox_cmd_load logic (which isn't called so frequently in comparison to string_pos I assume) to use it sounds like?

You can switch the input source to 'Clipboard' in Textractor; it should still pass the text to your websocket plugin, provided the plugin isn't configured to ignore clipboard data.

For some reason I thought the Clipboard page was just for viewing and didn't work the same as a hook input source. Sorry for my lack of knowledge here!

I just tried it, but it seems it doesn't work:

I confirmed the WebSocket server was running as expected first with the H-code option then switched to the Clipboard input source, but nothing from that point onward was sent via WebSocket.

The extension I'm using is this one: https://github.com/kuroahna/textractor_websocket

Around here is where it decides whether or not to broadcast the text in OnNewSentence:

• https://github.com/kuroahna/textractor_websocket/blob/7bbf74e84861da9bbd8c0afeee2dedd8b751257e/src/lib.rs#L48-L58
• get_current_select
• get_text_number

It doesn't seem to explicitly filter out the Clipboard TextNumber anywhere I can see, unless the UserSelectedTextThread stuff somehow makes it so the Clipboard is never seen as the selected thread before that or something?

Also tried it with https://github.com/KamWithK/TextractorSender (uses port 9001 instead and has some different JSON output structure) just to see, but it has the same result. It's checking even less before broadcasting here.

[Chenx221]

Oh, I see. So in theory I could increase my buffer setting, or I could find a hook for the gml_Script_textbox_cmd_load logic (which isn't called so frequently in comparison to string_pos I assume) to use it sounds like?

Increasing the buffer size won't completely solve the problem; you can't really guess how much "garbage content" the game engine might return.😂
If the game is compiled with YYC, hooking gml_Script_textbox_cmd_load becomes a more viable option.

I confirmed the WebSocket server was running as expected first with the H-code option then switched to the Clipboard input source, but nothing from that point onward was sent via WebSocket.

Regarding the WebSocket extension, you can try my previous build —just don't forget to remove the .xdll file with the same name, otherwise, it might not take effect.

[v2lmmj04]

Increasing the buffer size won't completely solve the problem; you can't really guess how much "garbage content" the game engine might return.😂
If the game is compiled with YYC, hooking gml_Script_textbox_cmd_load becomes a more viable option.

Time to take a loan out to buy enough RAM to handle all the garbage content with a huge buffer haha

I see, so kind of a trade-off of it being easier to mod when not compiled with YYC but easier to hook if it had been compiled.

Regarding the WebSocket extension, you can try my previous build —just don't forget to remove the .xdll file with the same name, otherwise, it might not take effect.

I just tried it, and it seems to work:

Do you have the source code for that anywhere? I'm curious how the code is different from the other one I typically use, for why this one works with the Clipboard input source but somehow the other one doesn't.

I did get a panic from it randomly just now when copying text for this reply, but can't seem to reproduce it (I thought maybe it was the emoji / some Unicode handling, but it works fine copying + sending that after restarting and trying that again so...). But, seems like I've got a full working solution now both for extracting the text + sending it over a WebSocket!

[Chenx221]

@v2lmmj04

Do you have the source code for that anywhere?

https://github.com/Chenx221/textractor_websocket

[v2lmmj04]

Ah, so TextNumber::TextThread(_) can never resolve to TextNumber::Clipboard then it seems based on that being the change needed. So there's additional handling needed to include Clipboard, rather than additional handling being needed to exclude it (which makes more sense for Console where you'd never really want it as an input source I guess)

Thanks again for all the help!