Kingdom of Deception [fan art, ai models, waiting for updates, future plans, healthcare]

[KGARrokh]

Spoiler: Lewds:

You don't have permission to view the spoiler content. Log in or register now.

Two more usefull resolutions:

ForgeUI	Source screenshot	Ratio
1312x738	1280x720	1,777
1170x878	512x384	1,333

Later i write C++ utility for automatching resolutions.

I wrote desmos calculator for this:

You must be registered to see the links

Input your width and height, scroll down to variants 1,2,3, one of this variants is your solution.
Because desmos doesn't support branching you need peek correct solution by yourself)

I could offer a bit of RenPy coding experience (basic) and a working SD_Forge Setup with a decent GPU.

I don't know and don't like python, but have some experience ("medium") in C/C++ (not modern C++) and GLSL (OpenGL, ShaderToy, but not VulkanAPI). I prefer complete remake of KOD, because i don't like in-game GUI and Log, both are garbage. But i don't want to make remake without developers permission, even those like Hreinn games; or another game (but KOD is more desireble for me).

Does Renpy allow to write pixel on screen (something like "write_pixel(w=0,h=0,RED)")? Or you can just show prerendered videos and images with this engine?

 

[zorin0815]

I don't know and don't like python,

For a simple VN with a few choices, you don't really need a lot of python coding besides defining some variables and some basic "if then" constructs.

I prefer complete remake of KOD, because i don't like in-game GUI and Log, both are garbage. But i don't want to make remake without developers permission

A full remake would be a lot of work for something we've already seen. I thought more of a fan made short side story in the KOD universe so it doesn't offend the original creators too much.
The "enthusiasm" in this forum will imho only last for a few dozend AI renders anyway.

Does Renpy allow to write pixel on screen (something like "write_pixel(w=0,h=0,RED)")? Or you can just show prerendered videos and images with this engine?

You can define stuff like boxes. But the main concept are full background images with multiple layers of foreground images which use transparent backgrounds

 

[zorin0815]

Foreign visitors in the Orc camp

Spoiler: Three heroines

You don't have permission to view the spoiler content. Log in or register now.

 

[Enigmanic]

Spoiler: Breeders for the Alliance

You don't have permission to view the spoiler content. Log in or register now.

 

[KGARrokh]

I found another AI hub, like CivitAI, which allows you to generate images and also use ADetailer, ControlNet, Upscale and e.t.c.
I already imported my KOD models from CivitAI. So if you don't have "good" GPU, you can try this site.

You must be registered to see the links

Pros:
txt2img;
img2img;
inpaint;
animate (look like only for SD 1.5);
ADetailer;
ControlNet;
Upscale;
NSFW allowed;
GUI similar to CivitAI (to be fair GUI is garbage, somehow even worse than CivitAI GUI);
free 100 credits per day (not much, but greater than 0).

Cons:
More restrictive in allowed settings (for example 25 sampler steps max for Free user);
Worse then CivitAI (in my experience, even with all Pros);
For some reason can't generate images with PonyDiffusionXL6 (just throw out an Exception and thats all). You need to use VAE;
Really dumb and useless search.

I still prefer CivitAI.

 

[KGARrokh]

For those who uses ForgeUI, you should read this:

You must be registered to see the links

In short, ForgeUI will become experimental branch, extensions workability doesn't guaranteed in future.
Developers recommend to switch to Automatic1111 (A1111).

I tested A1111 v1.9.3:
832x1216 image generated in 2:30 minutes with RTX 4080 Super

But don't worry) v1.9.3 is previous version.

Same task with v1.9.4: less then 10 seconds.

Now A1111 works pretty fast (v1.9.4).

You can still use ForgeUI, just don't update it (you can update extensions). And there is a matter of time when extension versions will become obsolete.

Spoiler: CivitAI generator (different concepts and styles):

You don't have permission to view the spoiler content. Log in or register now.

Spoiler: ForgeUI, ControlNet (Depth) pack 0:

You don't have permission to view the spoiler content. Log in or register now.

Spoiler: ForgeUI, ControlNet (Depth) pack 1:

You don't have permission to view the spoiler content. Log in or register now.

 

[Mamaragan]

a1111 work fine if you don't use anything. If you use shit like hiresfix, controlnet and adetailer, it becomes painfully slow. There's no reason whatsoever, for now, to change the UI. Forge is still blazing fast and support the extension that matter.

 

[zorin0815]

I tested A1111 v1.9.3:
832x1216 image generated in 2:30 minutes with RTX 4080 Super

You need to check you SD-Setup. An Image like this takes about 7 Seconds to generate on my 4090, which already includes the Adetailer (pure Image Render is 4 Seconds) -- and the 4090 should not be that much faster than your 4080.

I run SD WebUI Forge from the "Stability Matrix" Launcher with some enhanced Settings: --pin-shared-memoy, --cuda-malloc and --cuda-stream which speeds up the image generation quite a bit.

But even on my "vanilla" Automatic 1111 setup without these parameters a render like this does not take more than 15 Seconds.

 

[CourierNPC]

Several weeks' worth of trial and error with CivitAI

Spoiler

You don't have permission to view the spoiler content. Log in or register now.