Texture Replacement Mod from Natipiya (Including AI-Generated Content)(update 2026-04-20)

[knight77]

8 hours ago, Natipiya said:

只要看一下配置文件就可以发现，我只是用捕获到的贴图 hash值进行替换，配置文件大多结构都十分简单。事实上我也没搞清楚为什么会生效，我只是想使用您之前发布的match result ui时，试了各种如换d3dx.ini的操作之类的，然后就it just works了，我也期望有人能搞明白究竟是什么原理。

 

性能上我没进行过帧率的测试，对我自己似乎没什么明显影响，但可能不同设备上的区别会大一些。

 

为了贴近游戏整体风格我会做一些简单编辑，不过这方面我并不是什么行家，任何人想做一些改进或另起炉灶的话我都非常欢迎

 

If you take a look at the config files, you’ll see that I’m simply swapping out the captured texture hashes; the file structure is actually very straightforward. To be honest, I’m not entirely sure why it works. While trying to get your previously released Match Result UI to run, I experimented with various steps like swapping the d3dx.ini, and then—it just works. I’m honestly hoping someone else can figure out the exact logic behind it.

 

As for performance, I haven't done any formal FPS testing. It doesn't seem to have a noticeable impact on my end, though results might vary depending on the hardware.

 

I do some simple editing to try and match the game's overall aesthetic, but I'm no expert in that area. I’d be more than happy if anyone wants to make improvements or start a fresh version from scratch!

 

 

Someone else created the mod to replace the UI elements, but I cant find them in the forum here anymore. They might have left this community.

For the mod to work you need certain unique checks, since the hashes for a lot of the pictures/textures are the same, so hashes alone are not enough to exchange the textures in those cases.

 

I tried your mods and it didnt work for me.
I am on DMM though, but I assume your d3dx.ini file has certain code in it somewhere that does these unique checks.
I havent found them yet though in the d3dx.ini file you posted.

[knight77]

 

On 3/2/2026 at 4:41 PM, Natipiya said:

菜单立绘/Menu chibi

 

截至梅格

Up to Meg

 

快捷键/key: ctrl+t

半透明-全裸-原始/Semi-transparent - completely naked - primitive

菜单立绘.zip 4.15 MB · 2 downloads psd.zip 25.69 MB · 1 download

 

菜单小头像/Menu small chibi

 

截至梅格

Up to Meg

 

所有女孩都在一张贴图上，所以如果有新女孩或者你玩的是日服的话，mod会失效，需要重新调整。

All the girls are on the same texture, so if there are new girls or you're playing on the dmm server, the mod will not work and you'll need to readjust it.

 

快捷键/key: ctrl+shift+t

半透明-全裸-原始/Semi-transparent - completely naked - primitive

菜单头像.zip 6.43 MB · 2 downloads psd.zip 22.08 MB · 2 downloads

 

 

 

What did you use to modify the chibi girls?
I love it! Though its quite low resolution.
Did you use AI for this? This looks way better than what I tried doing by hand drawing.
I have higher resolutions for a lot of the girls since the developers sometimes post them on the official website.
Whatever you used, I would love to learn that to do it myself.

 

But in general if anyone is interested in editing the Chibis better, this is my current collection edited by hand:
Mega Download

 

Edited March 4 by knight77

[clxcst9527]

确实是生效不了，全下进mod文件夹只有一张温泉里的小春动漫图替换了，其他都没生效

[Natipiya]

25 minutes ago, knight77 said:

 

 

 

你用什么软件修改了这些 Q 版女孩？
我喜欢！虽然分辨率有点低。
这是用 AI 画的吗？这比我手绘的要好太多了。
我有很多女孩的高清图片，因为开发商有时会在官方网站上发布这些图片。
无论你用的是什么方法，我都非常想学会，这样我自己也能做到。

 

但总的来说，如果有人对更好地编辑 Q 版人物感兴趣，这是我目前手工编辑的合集：
Mega 下载

 

主要使用的是ai生成加上一部分简单的人工修改

 

ai生成使用的我主要使用comfyui，这是目前最主流的本地部署ai生图工具，特点是能自己布置工作流和适用各种本地模型，并且github上有大量拓展。另外一种主流工具是stable-diffusion-webui，这个不用自定义工作流，使用起来比较方便容易上手，也有大量拓展。

 

要在本地部署comfyui有多种方法，对于我来说会使用中国视频网站bilibili上up主秋葉aaaki分享的整合包，整合包会自动准备好前置环境并有一个能管理各种拓展更新的启动器。

 

安装好comfyui后，还需要准备工作流和各种模型，工作流可以自己搭建，但可能直接从别人分享的开始用起更加方便。

我以Aaalice大佬在其discord频道里面分享的工作流为基础，并自己做了一些修改，分成专门负责【文生图】、【图片局部重绘】等不同功能的工作流。当你导入别人的工作流时，会提示你缺失的拓展，可以用如下图的comfui manager安装所有缺失的拓展，如果你用的是启动器的话则应该在启动器里一个个手动安装

模型主要包括底模、lora模型、放大模型、细化检测模型等，所有模型基本都可以在civitai上面找到。

• 底模是基础模型，现在二次元画风常用illustrious模型作为底模。
• lora模型可以控制各种概念，包括角色、动作和画风等，图片局部重绘可能不需要lora，文生图时我会使用对应角色lora和自己炼制的游戏内chibi画风lora。lora模型必须与底模匹配。
• 其他模型你可以在工作流运行报错时，根据报错去找对应模型下载。

您引用的这部分主要使用图片局部重绘实现，具体步骤如下：

1. 裁剪出原图一部分，进行涂鸦：比如我会将女孩主体部分裁剪出来，先简单用肤色涂抹想要裸体化的部分，用粉色在乳房点上两点，用背景色涂抹多余部分
2. 导入负责【图片局部重绘】的工作流
3. 给要重绘部分涂上遮罩
4. 写一些提示词（如nude,nipples等）
5. 然后多次运行选择满意的结果，如果一直没有满意结果，可能选一张图片重回第一步
6. 将结果图片去除背景，跟游戏的原始图片放在一起编辑

如果要用文生图生成全新图片，我会用【文生图】工作流，根据想法搭配lora和提示词，生成比较满意的图像后再进入图片局部重绘步骤。

 

简言之，使用comfyui进行ai生成图像可以概括为以下步骤：

1. 在本地电脑部署comfyui，使用github上的comfyui库或者使用整合包
2. 自己做或者找一个工作流，安装缺失的拓展
3. 下载各种模型，解决各种报错
4. 根据你的想法搭配lora，编写提示词，图片重绘的话还要编辑和导入图像，然后就可以开始运行了

 

The process primarily utilizes AI generation supplemented by minor manual modifications.

Core Tools: ComfyUI vs. SD-WebUI

I primarily use ComfyUI, which is currently the most popular tool for locally deployed AI image generation. Its main advantages are the ability to build custom workflows, compatibility with various local models, and a vast library of extensions available on GitHub.

Another mainstream alternative is stable-diffusion-webui (SD-WebUI). It doesn’t require custom workflows, making it more user-friendly and easier to pick up, while still offering plenty of extensions.

Deployment and Setup

There are several ways to deploy ComfyUI locally. I personally use the one-click integration package shared by the uploader 秋葉aaaki on the Chinese video site Bilibili. This package automatically sets up the necessary environments and includes a launcher to manage extension updates.

Once ComfyUI is installed, you need to prepare workflows and models:

• Workflows: While you can build them from scratch, it is often more convenient to start with shared ones. I use workflows shared by Aaalice in their discord频道 channel as a foundation, which I have modified into specialized versions for Text-to-Image (T2I) and Inpainting.

• Managing Extensions: When importing a shared workflow, ComfyUI may notify you of missing nodes. You can use ComfyUI Manager to install them automatically. However, if you are using a specific launcher, you may need to install them manually one by one.

• 

Models

Models can generally be found on civitai. Key types include:

• Base Models: The foundation of the image. For anime styles, the Illustrious model is currently the standard choice.

• LoRA Models: Used to control specific concepts like characters, poses, or art styles. For Text-to-Image, I use specific character LoRAs and a custom-trained in-game chibi style LoRA. LoRA Models must match base models. LoRAs must match the architecture of the base model.

• Ancillary Models: This includes Upscalers and Refiner/Detection models. If the workflow throws an error, you can identify and download the missing model based on the error message.

---

Step-by-Step Process: Inpainting (Image Modification)

The specific example you referenced was achieved through Inpainting. The steps are as follows:

1. Crop and Sketch: Crop a section of the original image and perform a rough "doodle." For example, I crop the main body of a character, use skin tones to paint over areas for "nuking" (nudity), add pink dots for nipples, and use background colors to mask unwanted details.

2. Import Workflow: Load the specialized Inpainting workflow.

3. Masking: Draw a mask over the specific area to be regenerated.

4. Prompts: Input relevant tags (e.g., nude, nipples).

5. Iteration: Run the generation multiple times until a satisfactory result is achieved. If none are perfect, I may take a generated image and return to Step 1.

6. Final Edit: Remove the background from the result and composite it with the original game assets.

For creating entirely new images, I use the Text-to-Image workflow, combining ideas with specific LoRAs and prompts, then move to the Inpainting stage once a base image is selected.

---

Summary of the ComfyUI Workflow

1. Deployment: Install ComfyUI locally via GitHub or an integration package.

2. Workflow Setup: Build or download a workflow and install missing extensions.

3. Model Prep: Download necessary models and resolve any runtime errors.

4. Generation: Configure LoRAs and prompts. For Inpainting, edit and import your source images, then run the process.

 

这是我用来生成chibi画风图片的工作流，局部重绘部分也在其中，供你参考

This is the workflow I use to generate chibi-style images, which also includes the inpainting section. For your reference.

ZZZ-Aaalice的工作流_一键包_v12.2_正式版 (贴纸).zip

[Natipiya]

7 minutes ago, clxcst9527 said:

确实是生效不了，全下进mod文件夹只有一张温泉里的小春动漫图替换了，其他都没生效

温泉里的小春图不是替换的ui，所以应该会跟别的服装mod一样生效。

 

d3dx、分辨率都确认了吗？

不行的话我也不知道什么情况。d3dx对的话，放着不管可能哪天就开始生效了，这是认真的建议。

 

如果有其他人生效了的话，可以贴张图说一下吗？

If anyone else gets this mod working, could you post a screenshot and let me know?

[theshy1]

大佬·，我也只有小春温泉图生效

[Natipiya]

我似乎发现问题所在了。ui替换似乎靠的是krunkDunk大佬加载画面替换mod里的配置，大家可以试一下，用我发的d3dx并安装这个mod后ui替换是否生效

I think I’ve found the issue. The UI replacement seems to rely on the configuration from KrunkDunk's Loading Screen Replacement mod. Could everyone try using the d3dx file I provided alongside that mod and see if the UI replacement works?

 

更具体来说，起效的似乎是这一部分

More specifically, it seems that this part is the one that is effective.

 

[ShaderOverridePS_UI]

hash = f0eccd5df771e1a1

checktextureoverride = ps-t0

allow_duplicate_hash = overrule

Edited March 5 by Natipiya

[砂夜子]

DOA.7z

[Natipiya]

12 minutes ago, 砂夜子 said:

 I installed this mod, but it only took effect after I installed yours.And it reports an error.

 

 

18 minutes ago, 砂夜子 said:

 I installed this mod, but it only took effect after I installed yours.And it reports an error.

 

我也有这个报错，因为不影响使用（而且也搞不懂）所以就没管，如果有人能弄明白其中各种奥秘就好了。

至于你说的krunkDunk大佬的mod不起作用，可能是游戏hash值改变了，我自己添加了这两个hash值就能使用了，你也可以加入试试。另外我禁用了radioTransition.ini，因为从原帖里看到有人建议这么做

 

I get that same error too. Since it doesn't actually break anything (and I can't wrap my head around it anyway), I’ve just been ignoring it. It would be great if someone could eventually figure out the 'mysteries' behind it.

Regarding krunkDunk’s mod not working, it might be because the game's hashes have changed. I managed to get it working by adding these two hashes myself—you should give that a try. Also, I disabled radioTransition.ini because I saw someone in the original thread suggesting to do so.

 

[TextureOverrideLoadingScreen 1080prepaired]
hash = eaca3fcc
Preset = LoadingScene

 

[TextureOverrideLoadingScreen 2160prepaired]
hash = 90d4bbf8
Preset = LoadingScene
 

[knight77]

On 3/2/2026 at 3:02 PM, Natipiya said:

Specifically, this section of the code seems to be what makes it work; copy-pasting just this part should be sufficient.

[ShaderOverridePS_UI]
hash = f0eccd5df771e1a1
checktextureoverride = ps-t0
allow_duplicate_hash = overrule

That indeed seems to be the important addition!
You need to have that in an active mod for the others to work.
 

After some testing I think I also figured out which commands you need to have in the d3dx.ini file for it to work.

 

[CommandListHUD]
if $dump_ui
	dump = deferred_ctx_accurate share_dupes ps-t0 mono dds
   	analyse_options = deferred_ctx_accurate share_dupes dump_vb txt buf
   	dump = deferred_ctx_accurate ps-t0 txt buf
endif
if !$hud_visible
	handling = skip
endif

 

I needed to add the lines:
 

   	analyse_options = deferred_ctx_accurate share_dupes dump_vb txt buf
   	dump = deferred_ctx_accurate ps-t0 txt buf

 

But also I needed to switch:

track_texture_updates=2

 to:

track_texture_updates=1

 

And all of this only worked after game restart. Updating with F10 did not work!

I am testing more currently. But so far, these changes helped with the "Match Result Mod" to work with your version.
Now I try to get others working as well. Just need different hashes in some cases.

[Natipiya]

27 minutes ago, knight77 said:

I am testing more currently. But so far, these changes helped with the "Match Result Mod" to work with your version.
Now I try to get others working as well. Just need different hashes in some cases.

感谢你的辛勤测试，我将您的测试成果放在开头了，以便更多人能看见

Thank you for your hard work on these tests! I’ve put your findings to the beginning of the post so that more people can see them.

[knight77]

3 hours ago, Natipiya said:

感谢你的辛勤测试，我将您的测试成果放在开头了，以便更多人能看见

Thank you for your hard work on these tests! I’ve put your findings to the beginning of the post so that more people can see them.

 

Ok, you know that I play on DMM, right?
Just so you know.
With these changes it works on DMM as well, but need some updated hashes in some cases. For the girls for example, since we are at 33 girls in DMM version now.

 

3 hours ago, 砂夜子 said:

Can you create mods for these two swimsuits? These are the only two SSR swimsuits on Steam that don't have corresponding mods yet.

 

 

First off, I am on DMM Player and second, these suits are covering quite the full body, which means the body undearneath is basically non-existent, which doesnt offer much to mod there.

Plus this thread is about modding texture or more precisely the UI elements of the game.

Edited March 5 by knight77

[knight77]

It WORKS! And I LOVE IT!

Just tested it with updated Ayane hashes:

Spoiler

 

 

Though there is sooo much work to do now...

 

I do like a lot of of your chibi creations, though personally I would like to have more creative controle of the images.
You already gave some explaination about the AI image generation process, but that is completely new for me, so that will take some time to understand or get started with.

[knight77]

On 3/2/2026 at 6:27 PM, Natipiya said:

挑战赛界面/Challenge interface

 

对挑战赛界面许多贴图都进行了修改，从大地图到结算界面

Many textures on the challenge interface have been modified, from the world map to the results screen.

 

对大地图的修改只是半成品，将酒店变成情趣酒店，其他地方要如何修改没有什么思绪

The modifications to the large map are only half-finished; turning hotels into love hotels is one thing, but there's no clear idea how to modify other areas.

 

 

 

 

 

That girl is missing from the PSD files.
Could you share that as well, please?

[Natipiya]

8 hours ago, knight77 said:

 

That girl is missing from the PSD files.
Could you share that as well, please?

这个psd文件没有被保存，不过我更新了文件，加入了它的原始png图片

The PSD file wasn't saved, but I’ve updated the files to include the original PNG images.

 

12 hours ago, knight77 said:

You already gave some explaination about the AI image generation process, but that is completely new for me, so that will take some time to understand or get started with.

comfyui只要完成部署，使用现成工作流并不复杂，应该很快就能上手，特别对于有美术功底的人来说，修复生成结果应该也会更轻松

Once ComfyUI is deployed, using existing workflows isn't that complicated—you should be able to get the hang of it pretty quickly. Especially with your background in art, fixing the generated results should be even easier for you.

[QQAASS]

很不错的模组！温泉的模组文件好像缺少了一些

[Natipiya]

1 hour ago, QQAASS said:

很不错的模组！温泉的模组文件好像缺少了一些

 

抱歉，已经补上了。另外像这种的整张图片，也可以简单地换成自己喜欢的图片

[QQAASS]

我尝试自己添加了一些动态视频，效果还不错屏幕录制 2026-03-08 .zip。krunkDunk大佬加载画面替换mod不知道什么原因很容易崩溃，我只能复制一小部分代码使用，确保这个模组能运行

[QQAASS]

2 minutes ago, 砂夜子 said:

我想用我的GIF来加载模组。

我是用这个主题里的主页替换动态图的方法替换的，我没有想过替换加载界面，因为加载的时候我都是后台等待，这个游戏的服务器太小了。

[Natipiya]

启动器/launcher

 

素材来源/Source of materials: https://www.deviantart.com/loveeuphoria/art/24-girls-nude-version-1289726997

 

本mod贴图带中文，其他语言需要翻译和重新捕获hash值。

This mod includes Chinese textures; textures in other languages require translation and recapture of hash values.

 

这个mod使用了不同的着色器，需要更新d3dx.ini文件，详细说明看下一楼，更新后文件看一楼

This mod uses different shaders and requires an update to the d3dx.ini file. Please check the next post for detailed instructions, and find the updated file in the first post.

启动器.zip8f908be2-B8G8R8A8_UNORM.zip

Edited March 9 by Natipiya

[Natipiya]

在d3dx中添加ui贴图对应的vs着色器ShaderOverride，可以在hunting ui 时捕捉到最底层的贴图。

如启动器界面的vs着色器7fba8a96fcba735d，添加成[ShaderOverrideHUD4]后，再hunting ui就能在"游戏根目录\FrameAnalysisDeduped"下找到没有按钮的背景图，使用“8f908be2”这个hash值就不会与其他贴图hash重复。 

Adding a ShaderOverride for the UI texture's VS (Vertex Shader) in d3dx.ini allows you to capture the bottom-most texture layer during UI hunting.

For example, by adding the launcher interface's VS shader 7fba8a96fcba735d as [ShaderOverrideHUD4], you can then hunt the UI to find the button-less background image in the \FrameAnalysisDeduped folder. Using the hash value 8f908be2 ensures there are no conflicts with other texture hashes.

而要使得ui替换生效，还要有对应ps着色器的ShaderOverride，在上面例子中，就是ps着色器e9c1db417d2b3b58。在完全退出游戏的前提下写入[ShaderOverridePS_UI3]，变更可能才会生效

To make the UI replacement take effect, you also need a ShaderOverride for the corresponding PS (Pixel Shader)—in this case, PS shader e9c1db417d2b3b58. Note that you may need to completely exit the game before writing [ShaderOverridePS_UI3] for the changes to apply.

[Ghosts371]

Krunk's loading screen mod with the lewd bouncing honoka was working but after I tried the new d3dx.ini from here  

the lewd bouncing honoka stopped working.  my d3dx.ini does have these changes 

[CommandListHUD]
if $dump_ui
    dump = deferred_ctx_accurate share_dupes ps-t0 mono dds
       analyse_options = deferred_ctx_accurate share_dupes dump_vb txt buf
       dump = deferred_ctx_accurate ps-t0 txt buf
endif
if !$hud_visible
    handling = skip
endif

 

Additionally, setting the value of track_texture_updates to 1" and "

[ShaderOverridePS_UI]

hash = f0eccd5df771e1a1

checktextureoverride = ps-t0

allow_duplicate_hash = overrule"  and I tried deleting the Shader cache folder and the d3dx.user.ini he loading screen mod with the lewd honoka still won't work anymore. I clicking the options in the GUI but still nothing. Does anyone know how to enable the lewd bouncing honoka from the loading screen mod while enabling the other 2d image chibi replacement mods to work?

[Natipiya]

 

1 hour ago, Ghosts371 said:

Krunk's loading screen mod with the lewd bouncing honoka was working but after I tried the new d3dx.ini from here  

我这边可以两个一起使用。
加载界面的图像替换生效吗？ 如果加载界面图像也没有替换，或许是加载界面mod的hash值不匹配。可以试试在这个mod里面添加

I can use both of them together on my end.

Does the image replacement for the loading screen work?

If the loading screen image isn't being replaced either, there might be a hash mismatch in the loading screen mod. You can try adding the following lines to the mod configuration:

 

[TextureOverrideLoadingScreen 1080prepaired]
hash = eaca3fcc
Preset = LoadingScene

[TextureOverrideLoadingScreen 2160prepaired]
hash = 90d4bbf8
Preset = LoadingScene

或者hunting你的游戏中的这张背景图片，看看hash值是否存在配置文件中

Alternatively, try hunting the background image in your game to see if the captured hash value matches the one in the configuration file.


如果还是不行的话，将你的d3dx.ini和LoadingScreenMod的main.ini打包发出来看看？

If it’s still not working, could you please package your d3dx.ini and the main.ini from the LoadingScreenMod and send them over for me to take a look?

[Ghosts371]

I really appreciate you responding, thank you
I am using steam, not sure if that makes a difference or not
The image replacement was working when I was using knight 77's d3dx.ini from here 

https://www.loverslab.com/topic/263501-texture-replacement-mod-from-natipiya-including-ai-generated-content

but once I switched to the d3dx.ini from here: https://www.loverslab.com/topic/263501-texture-replacement-mod-from-natipiya-including-ai-generated-content/
then it stopped (I also just tried making the changes in knight 77's d3dx.ini but the image replacement/honoka still stopped working)

You said to "You can try adding the following lines to the mod configuration:

 

[TextureOverrideLoadingScreen 1080prepaired]
hash = eaca3fcc
Preset = LoadingScene

 

[TextureOverrideLoadingScreen 2160prepaired]
hash = 90d4bbf8
Preset = LoadingScene"

 

are you saying to edit the "main.ini" in the folder "_00LoadingScreenMod" in the loading screen mod to add those? Would I put them under the section that says:

 

"; Hide Mod when detect hash from notification screen
[TextureOverride Hash 4]
hash = d1638eba
match_priority = -1
Preset = HideMod"

Edited March 11 by Ghosts371

[Natipiya]

2 hours ago, Ghosts371 said:

are you saying to edit the "main.ini" in the folder "_00LoadingScreenMod" in the loading screen mod to add those? Would I put them under the section that says:

 

对。在ini中的具体位置应该不影响，不过我会放在这些后面

Yes. The specific location in the ini file shouldn't matter, but I'll put it after these.