Another merge from me. Not so playful, more classic facial features.
As always, I paid attention to quality and did a lot of tests before I published it. Sometimes the merge is a bit buggy with the eyes, as a remedy just add a lora character model with a low value.
If you like my work, feel free to leave a tip here ;)
Hey,
I built a small tool called PromptCleaner – it removes brackets and weightings like (word:1.2) from AI prompts.
Maybe it’s useful for someone. Works with things like Stable Diffusion, ComfyUI, and similar tools.
👉 https://github.com/LordKa-Berlin/PromptCleaner
And as always with me
ATTENTION
for all who want to recreate my work!
I use the Dynamic Thresholding Extension for Stable Diffusion for almost all my work
https://github.com/mcmonkeyprojects/sd-dynamic-thresholding
This makes higher CFG values and steps possible!
I can only recommend it to you!
Here is a Lora that I can absolutely recommend, I am very happy to have found it because it can create fantastic lighting moods!
Description
FAQ
Comments (7)
You have outdone yourself. Great model.
Many thanks for the praise! I'm glad if you like it.
nice model, but you know with the excessive prompt you use half of it is not even read by clip encoder there to long
The most important thing for me is that the things in which I describe what it should do, how it should look etc. work, it does, even if it is at the very end of the prompt. The other thing is the text modules (styles) that I use “over and over again”, which I use as required; I don't care how much of it the encoder reads, the overall result is then decisive. If, for example, I delete something, even if it's a small thing, the overall result can change fundamentally because, for example, the order or something else changes. As I said, I have a predefined style, which I then “only” add the image properties to.
So far, it's working quite well for me, I think ;)
Many thanks for the praise and your advice!
@LordKa agree with the OP, you can DEFINITELY delete 90% of your prompt bc Pony doesn't understand these tags anymore (SDXL did). The effects you're seeing are placebo or come from prompt order and relative weighting, Still, the model looks nice and I'll try it!
Best checkpoint hands down
Thank you very much!
Details
Files
Available On (3 platforms)
Same model published on other platforms. May have additional downloads or version variants.