Mod Lab Rats 2 Reformulate GenAI Mod Thread

[gimb]

does this mod still only use outdated A1111?

also does it function better with illustrious checkpoints yet?

sorry if these questions were answered here or the game thread, its been months since i looked at the status of this mod.

Afaik there is currently no mod that works with comfyui, however since almost everything else uses the same api calls most other frameworks like forge,forge classic, sd.next and maybe even swarmui should work.

For your question if illustrious Checkpoints work i'd reccomend the original implementation of maswoo with it's json prompts

You must be registered to see the links

there are three json for an illustrious model in there which you can use as base.

 

[lynxis]

does this mod still only use outdated A1111?

also does it function better with illustrious checkpoints yet?

sorry if these questions were answered here or the game thread, its been months since i looked at the status of this mod.

ComfyUI fork by trojal from the LR2 discord

You must be registered to see the links

Free Cities pregmod has a very mature genAI implementation for Comfy if anyone wants to look at their git for inspiration

 

[elpe]

I'm trying right now to get busoni position mod (did not get to run it in your og version) and and gimb background mod (works very well) to work together. I'm hitting a wall here maybe someone with more skill can help.

Inbeteen I made some promtlists for different hantai syle lists that work good to well. So I can give back. All are WiP!
I choose a random seed for the example image so it's not ideal but the same for all images. Play around a bit.

 

[gimb]

I'm trying right now to get busoni position mod (did not get to run it in your og version) and and gimb background mod (works very well) to work together. I'm hitting a wall here maybe someone with more skill can help.

Inbeteen I made some promtlists for different hantai syle lists that work good to well. So I can give back. All are WiP!
I choose a random seed for the example image so it's not ideal but the same for all images. Play around a bit.

Whats the thing that's not working?

Some hints for my Version one of the major changes is adding img2img to the json and latent modifier integrated (sadly only works with forge) to img2img. If you enable it you can play around with a high img2img de-noising strength and the extra noise multiplier in latent modifier to get better pose and clothing adherence. Currently i'm thinking of modifying the code a little bit and adding latent modifier as an extra optionin the json independent of img2img.

Some examples in order no latent modifier no img2img, latent modifier no img2img, both latent modifer and img2img

 

[elpe]

Whats the thing that's not working?

Some hints for my Version one of the major changes is adding img2img to the json and latent modifier integrated (sadly only works with forge) to img2img. If you enable it you can play around with a high img2img de-noising strength and the extra noise multiplier in latent modifier to get better pose and clothing adherence. Currently i'm thinking of modifying the code a little bit and adding latent modifier as an extra optionin the json independent of img2img.

Some examples in order no latent modifier no img2img, latent modifier no img2img, both latent modifer and img2img

Your stuff works fine it's the base for the merge. I was just to optimistic I could just merge the files by editing the GenAi files and moving the context_auto_cleaner, position_pose_hooks and promt_enhancer in. I worked out some flaws but others stay. I can not explain that much because I don't realy know what I doing to be honest ^-.-

Maybe it's just the promt file now that is wrong.

 

[zalamander]

Ok, it works for me, as long as i only have ONE of the following installed:

- Stable Difussion Forge, from iilyasviel (

You must be registered to see the links

)
- Automatic1111: (

You must be registered to see the links

)
- Stable Difussion Next: (

You must be registered to see the links

)

That's because Automatic1111 only build using python 3.10.6 at this date (will give an error the first time the installer in run, just tun the second time to bypass the error), Stable Difusion Forge needs Python 3.11.9, and SD.Next dependencies build properly under Python 3.12.10 and up. The other versions waaazzzaaa mentions have subtle differencies in their api's that requires some manual patching to make them work. (No fork of stable diffusion has a stable API, mind you, and everyone seems to think that their versions are better than the rest ¯\_(ツ)_/¯). And realistically, since i own an Radeon 6900XT Automatic1111 fills the bill for me (I tested the others, some build, some run, some are problematic, and you know, time is limited). To support other versions should be rather easy if you get the way to detect how the software identifies themselves via API, and what changes requires to make things work, but that's work for somebody else.).

 

[Selek]

I don't suppose this mod will ever support video, will it? I've ben playing around with Meta AI's free video generator, and it comes up with impressive results. Is there a NSFW version of the Meta AI thing?

 

[busoni]

I'm trying right now to get busoni position mod (did not get to run it in your og version) and and gimb background mod (works very well) to work together. I'm hitting a wall here maybe someone with more skill can help.

Inbeteen I made some promtlists for different hantai syle lists that work good to well. So I can give back. All are WiP!
I choose a random seed for the example image so it's not ideal but the same for all images. Play around a bit.

Sorry, I never intended to make "a new version" of the mod, just made changes based on what I wanted from it. I included a location checker to include it on the rompt and make the image background relevant to the location (office, bedroom, etc). Not sure if that clashes with what you say about the backgrounds from the other version.

 

[gimb]

I don't suppose this mod will ever support video, will it? I've ben playing around with Meta AI's free video generator, and it comes up with impressive results. Is there a NSFW version of the Meta AI thing?

Just browse Civitai, there are enough video generators out there. However the generation time will be quite long even with the newest hardware.

 

[BackRunner]

This looks great and all..... But I am too stupid and all of this are going over my head.

 

[whatifjoanwasoneofus]

does this mod still only use outdated A1111?

I use forge instead of A1111, it runs much faster. ~10 sec DMD2 SDXL 1024x images. 4060 8GB Vram

 

[gimb]

Your stuff works fine it's the base for the merge. I was just to optimistic I could just merge the files by editing the GenAi files and moving the context_auto_cleaner, position_pose_hooks and promt_enhancer in. I worked out some flaws but others stay. I can not explain that much because I don't realy know what I doing to be honest ^-.-

Maybe it's just the promt file now that is wrong.

Here is my new version all previous changes and all Sex positions in game\sex_positions\_position_definitions_ren should now be supported by the json prompt (if added to the prompt file see DMD2_Working) .For the Sex positions to work you need to be on the newest Version, and the first few generated pictures for a person might be in the wrong position.

 

[Carmen Sandiego]

I use forge instead of A1111, it runs much faster. ~10 sec DMD2 SDXL 1024x images. 4060 8GB Vram

I can't get forge to do shit. SD is allegedly setup, mod "works" in game. I click to generate image, it says "generating" for like a second then does nothing. I look at the bat or whatever and SD either isn't getting any info from the game ort it is and just isn't generating images. I only tried because your comment said it was faster, but the regular SD works better, or at all, for me.

 

[gimb]

I can't get forge to do shit. SD is allegedly setup, mod "works" in game. I click to generate image, it says "generating" for like a second then does nothing. I look at the bat or whatever and SD either isn't getting any info from the game ort it is and just isn't generating images. I only tried because your comment said it was faster, but the regular SD works better, or at all, for me.

Are you using the version from the op or a different version? The one from Op should say generating 0% if it doesn't work.

For your issue you can test if you can generate pictures at all with the normal webGUI

You must be registered to see the links

if that works check the console and the log file in the game folder.

For me setting up forge was the same as a111
and it really is faster

Also if you have an AMD graphics card it might only work with a111

 

[Carmen Sandiego]

Are you using the version from the op or a different version? The one from Op should say generating 0% if it doesn't work.

For your issue you can test if you can generate pictures at all with the normal webGUI

You must be registered to see the links

if that works check the console and the log file in the game folder.

For me setting up forge was the same as a111
and it really is faster

Also if you have an AMD graphics card it might only work with a111

Honestly, I'm not sure... I downloaded like 3 to see if I could get any to work but I think I started in reverse ordered and none of the initial 3 I tried worked. Ultimately I saw somebody pose something about stability matrix and I decided to try that to make shit easier and used I the first ai mod, which worked. I also tried the other three I initially tried, those didn't work so I guess those just don't work? Anyway, now that I have a working point I'll see about optimizing. Not sure if forge is faster for me, but I'll definitely run some tests to see. Fun stuffs to play around with.

 

[Arasutaru]

anyone have that issue where the refresh and image controller covers the images?

 

[Magister Crudi]

anyone have that issue where the refresh and image controller covers the images?

Three ways to deal with that, depending on your hardware, which version of GenAI you use and what you prefer.

1. If you're not already running the game at your maximum screen resolution, change it.

2. If your GenAI config menu lets you adjust image resolution, fiddle with it until you get something you prefer.

3. Edit your GenAI source to move the image UI. The following example is for AnotherMike's mods.

Go to your GenAI mod folder.
Open SD_UI.rpy with a text editor like Notepad.
Find "screen SD_update_button()" (without the quotes).
Scroll down to "yalign 0.2" and change it to "yalign 0.13". (You may need to fiddle the values a few times until it ends up where you want it. Decreasing yalign moves the controls up the screen. Increasing xalign moves them to the right.)
Save and exit.
After loading the game, use 'shift-R' to force a rebuild from the new source.

 

[Yllarius]

Here is my new version all previous changes and all Sex positions in game\sex_positions\_position_definitions_ren should now be supported by the json prompt (if added to the prompt file see DMD2_Working) .For the Sex positions to work you need to be on the newest Version, and the first few generated pictures for a person might be in the wrong position.

I can't seem to get Appends to work with this at all. Like, I add in additional positive or negative prompts and nothing changes. I've even tried deleting the images and rerolling the seed and still nothing

 

[x760917]

Edit this file

SD_UI.rpy

423
screen SD_update_button():
zorder 11
layer "solo"

frame:
xalign 0.95
yalign 0
background "#1a45a1aa"
padding (20, 20) # Padding inside the frame

516
screen generating_indicator():
zorder 101 # Ensure it's above the image but below the fullscreen overlay potentially
layer "solo"

frame:
background "#000000aa" # Semi-transparent background
xalign 0.925
yalign 0.1

Effect

 

[Arasutaru]

Edit this file

SD_UI.rpy

423
screen SD_update_button():
zorder 11
layer "solo"

frame:
xalign 0.95
yalign 0
background "#1a45a1aa"
padding (20, 20) # Padding inside the frame

516
screen generating_indicator():
zorder 101 # Ensure it's above the image but below the fullscreen overlay potentially
layer "solo"

frame:
background "#000000aa" # Semi-transparent background
xalign 0.925
yalign 0.1

Effect


View attachment 5377353 View attachment 5377360

hold up how'd you cut out the background? i get a big ass white background and nothing else for whatever reason