Hello everyone!In the course of playing this game, I realized that normal girl packs usually suffer from inconsistent picture styles and missing pictures of a certain tag. It just so happens that I also often use novel ai to make some anime girl pictures. So I tried to make a girl pack with novel ai. I think by replacing the character name and lora model, it's easy to make a girl pack of other girls that you want.
My exact method is:
    1. make as many pictures as I wanted using the usual method of generating pictures for a specific girl. Whenever I generate the image I want, I save the image and record the tags used in a text file.
  2. When I'm done, I save this text file.
  3. When I need to make other girl packs, I use the replace function to replace the name of the original girl in the text file with the name of another girl, and replace the lora model of the original girl with the lora model of another girl.
  4. Use the replaced tags to run in novelai, you can generate a unified style of pictures, and use this to create a girl package.
I made a girl pack first and uploaded it, the girl is Suzuran from Tomorrow's Ark. If you think it's acceptable to use ai painting to make a girl pack, and need more detailed methods, please leave a comment under the post. If you think the girl packs made by ai painting are unacceptable, please let me know and I'll quit uploading more girl packs made by ai painting.
I'm a Chinese player and not very good at English, so I'm using a translator to write this post, and I apologize for any grammatical problems.


download link:https://mega.nz/file/tbNDXBJZ#52syqX63bca4c5jieG-ua7dfyyio730xqs4K1SEZZcY

Quote from: peacedream on Dec 28, 2023, 04:52 PMHello everyone!In the course of playing this game, I realized that normal girl packs usually suffer from inconsistent picture styles and missing pictures of a certain tag. It just so happens that I also often use novel ai to make some anime girl pictures. So I tried to make a girl pack with novel ai. I think by replacing the character name and lora model, it's easy to make a girl pack of other girls that you want.
My exact method is:
    1. make as many pictures as I wanted using the usual method of generating pictures for a specific girl. Whenever I generate the image I want, I save the image and record the tags used in a text file.
  2. When I'm done, I save this text file.
  3. When I need to make other girl packs, I use the replace function to replace the name of the original girl in the text file with the name of another girl, and replace the lora model of the original girl with the lora model of another girl.
  4. Use the replaced tags to run in novelai, you can generate a unified style of pictures, and use this to create a girl package.
I made a girl pack first and uploaded it, the girl is Suzuran from Tomorrow's Ark. If you think it's acceptable to use ai painting to make a girl pack, and need more detailed methods, please leave a comment under the post. If you think the girl packs made by ai painting are unacceptable, please let me know and I'll quit uploading more girl packs made by ai painting.
I'm a Chinese player and not very good at English, so I'm using a translator to write this post, and I apologize for any grammatical problems.


download link:https://mega.nz/file/tbNDXBJZ#52syqX63bca4c5jieG-ua7dfyyio730xqs4K1SEZZcY

Interesting. How do you use Lora with NovelAI? Do you know of a guide for that?

Quote from: Goldo on Jan 05, 2024, 04:58 PMHow do you use Lora with NovelAI? Do you know of a guide for that?

On this site you can find a guide how you can train your Lora.

I use a stable diffusion integration pack made by a Chinese blogger. Simply download the required lora models from the civitai website and then place the models in the specified folder to use the models.
The stable diffusion integration package is at https://www.bilibili.com/video/BV1iM4y1y7oA/

Quote from: Goldo on Jan 05, 2024, 04:58 PM

Quote from: peacedream on Dec 28, 2023, 04:52 PMHello everyone!In the course of playing this game, I realized that normal girl packs usually suffer from inconsistent picture styles and missing pictures of a certain tag. It just so happens that I also often use novel ai to make some anime girl pictures. So I tried to make a girl pack with novel ai. I think by replacing the character name and lora model, it's easy to make a girl pack of other girls that you want.
My exact method is:
    1. make as many pictures as I wanted using the usual method of generating pictures for a specific girl. Whenever I generate the image I want, I save the image and record the tags used in a text file.
  2. When I'm done, I save this text file.
  3. When I need to make other girl packs, I use the replace function to replace the name of the original girl in the text file with the name of another girl, and replace the lora model of the original girl with the lora model of another girl.
  4. Use the replaced tags to run in novelai, you can generate a unified style of pictures, and use this to create a girl package.
I made a girl pack first and uploaded it, the girl is Suzuran from Tomorrow's Ark. If you think it's acceptable to use ai painting to make a girl pack, and need more detailed methods, please leave a comment under the post. If you think the girl packs made by ai painting are unacceptable, please let me know and I'll quit uploading more girl packs made by ai painting.
I'm a Chinese player and not very good at English, so I'm using a translator to write this post, and I apologize for any grammatical problems.


download link:https://mega.nz/file/tbNDXBJZ#52syqX63bca4c5jieG-ua7dfyyio730xqs4K1SEZZcY

Interesting. How do you use Lora with NovelAI? Do you know of a guide for that?

This video author has also produced tutorials on the use of lora, you can look through his drop-in videos

Cant figure out how Lora works but see a lotta potential. heres a untagged slime girl pack as conception etc:
https://mega.nz/file/MLFQxYrb#MjWWq8gxpngf8nqZ9RJ3m8ltGZQdSD49dpfCzyq-SGw

messed around witha  few models prompts etc

requests and suggestions welcome

Really can't go down that particular rabbit hole for now or I might never get to work on BK, but thank for the sources, I'll make sure to check them out later!

Hello everyone, I'm also a gamer from China, and like the original poster, I specifically found this website to share.

My English isn't great, so I used ChatGPT for translation, hoping everyone can understand.

But what I want to recommend is Stable Diffusion, especially the Pony XL model, which can be directly found on Civitai.

It can generate images of some well-known characters directly without overlaying Lora, which can be used to make girl packs. Because its technology is based on SDXL, the expressive power of image quality is quite good.

More importantly, if a set of workflows is developed for each scene generated, using ComfyUI for production, it can achieve high efficiency while ensuring image quality and art style, just by replacing the character names.

When I have time, I will try to make some girl packs and workflows to share as reference.

Someone know what AI ( or art style) are people using for images like this:
NSFW: https://kemono.su/patreon/user/117734244/post/101728296

Quote from: kingthandra on Apr 28, 2024, 06:43 PMSomeone know what AI ( or art style) are people using for images like this:
NSFW: https://kemono.su/patreon/user/117734244/post/101728296

I believe stable diffusion can accomplish this, especially with the Pony Diffusion model. Here's an example I generated featuring Tifa, a scene containing multiple characters intertwined with characters from other series. Lora merely used a few general prompts without employing specific ones for particular poses, character identities, or object references.

Currently, there are two local deployments of stable diffusion, and ComfyUI can directly access the workflow for creating images using stable diffusion.

https://mega.nz/file/EBYEmAQB#ovER-v1X0iGK3Wp6ZQGslwMXdu_TJrUvcce7RVVkHtA

Quote from: kingthandra on Apr 28, 2024, 06:43 PMSomeone know what AI ( or art style) are people using for images like this:
NSFW: https://kemono.su/patreon/user/117734244/post/101728296

I would recommend creating an account on civit.ai (you need an account to see NSFW posts) and browsing the image gallery there for a similar artstyle. Usually all metadata is included there, so you know exactly which model/checkpoint, lora, prompt, settings, seed, etc were used to create the image.

Quote from: Luck.zq on Apr 30, 2024, 07:12 AM

Quote from: kingthandra on Apr 28, 2024, 06:43 PMSomeone know what AI ( or art style) are people using for images like this:
NSFW: https://kemono.su/patreon/user/117734244/post/101728296

I believe stable diffusion can accomplish this, especially with the Pony Diffusion model. Here's an example I generated featuring Tifa, a scene containing multiple characters intertwined with characters from other series. Lora merely used a few general prompts without employing specific ones for particular poses, character identities, or object references.

Currently, there are two local deployments of stable diffusion, and ComfyUI can directly access the workflow for creating images using stable diffusion.

https://mega.nz/file/EBYEmAQB#ovER-v1X0iGK3Wp6ZQGslwMXdu_TJrUvcce7RVVkHtA

Looks like can be Pony Diffusion, this image is very close: https://civitai.com/images/5293520 (the style i mean)

I am really illiterate in AI, but I would like to learn.

Quote from: Goldo on Mar 04, 2024, 11:14 PMReally can't go down that particular rabbit hole for now or I might never get to work on BK, but thank for the sources, I'll make sure to check them out later!

*Narrator's voice*
He did in fact go down that particular rabbit hole. Many hours later, not much came out of it but at least he has some new item icons to replace the crappy pixelated ones that were in the game.

Quote from: kingthandra on Apr 30, 2024, 05:02 PM

Quote from: Luck.zq on Apr 30, 2024, 07:12 AM

Quote from: kingthandra on Apr 28, 2024, 06:43 PMSomeone know what AI ( or art style) are people using for images like this:
NSFW: https://kemono.su/patreon/user/117734244/post/101728296

I believe stable diffusion can accomplish this, especially with the Pony Diffusion model. Here's an example I generated featuring Tifa, a scene containing multiple characters intertwined with characters from other series. Lora merely used a few general prompts without employing specific ones for particular poses, character identities, or object references.

Currently, there are two local deployments of stable diffusion, and ComfyUI can directly access the workflow for creating images using stable diffusion.

https://mega.nz/file/EBYEmAQB#ovER-v1X0iGK3Wp6ZQGslwMXdu_TJrUvcce7RVVkHtA

Looks like can be Pony Diffusion, this image is very close: https://civitai.com/images/5293520 (the style i mean)

I am really illiterate in AI, but I would like to learn.

Actually, it's quite simple. Go for it.

Quote from: Goldo on Apr 30, 2024, 07:43 PM

Quote from: Goldo on Mar 04, 2024, 11:14 PMReally can't go down that particular rabbit hole for now or I might never get to work on BK, but thank for the sources, I'll make sure to check them out later!

*Narrator's voice*
He did in fact go down that particular rabbit hole. Many hours later, not much came out of it but at least he has some new item icons to replace the crappy pixelated ones that were in the game.

If you need images for item icons or a similar purpose, I highly recommend the Lora models created by Navimixu and ConceptConnoisseur, as they offer a wide variety of options. Here are their homepage links:
[Navimixu's Homepage](https://civitai.com/user/navimixu/models)
[ConceptConnoisseur's Homepage](https://civitai.com/user/ConceptConnoisseur/models)

Furthermore, if you're too occupied to create them and short on time, perhaps consider posting your requirements in a thread to gather contributions from others? Although I cannot guarantee excellent results, I'd be more than happy to assist by creating some of the images myself. Let's collaborate and see what we can achieve collectively.

Still relying on tools for the translations, I wonder if they're comprehensible, but my hope is that they serve their purpose.