Not Artists Styles for Pony Diffusion V6 XL
CLIP SKIP in ComfyUI
-1 gray picture with noise and spots
-2 picture looks the same as in the preview
If you do not add the CLIP Set Last Layer node, the result will be exactly the same as -2.
(It will not work correctly with other SDXL checkpoints.)
Does not work with any other models!
except Pony Diffusion V6 XL or models with Pony Diffusion V6 XL in the merge!
You can of course use with other models you no one forbids it, but I do not guarantee a good result!
LoRAs without Trigger Words! but there are words used in training!
Tokens that can enhance the style or change it!
Digital Art - digital art, realistic,
Concept Art/Concept Art Twilight - concept art, realistic, & concept art, horror (theme), monochrome, fog, realistic,
Oil Gothic - painting (medium), traditional media, realistic,
Gothic Art - painting (medium), traditional media, realistic,
Line Art - lineart, monochrome, greyscale,
Oil Painting - traditional media, realistic,
Photo - RAW, photo, realistic,
The previews were made without using ADetailer
TRAINED
Smooth, Anime, Ink, Oil Painting, Photo, Line Art, Summer Days, Gothic Art, Concept Art ,Smooth 2, Anime 2, Digital Art
MERGE
Smooth Anime - Smooth + Anime
Smooth Anime 2 - Smooth 2 + Anime 2
Oil Gothic - Oil Painting + Gothic Art
Concept Art Twilight - Concept Art + Concept Art Dark + Concept Art Negative
Styles are based partially or entirely on images created by the AI
---------------------------------------------------------------------------------------------------------------------
Type
|
LORA
|
Stats |
16271
0
|
Published |
2024-04-14 12:03:13 |
Base Model |
Pony
|
Usage Tips |
Clip Skip: 2
|
Trigger Words |
concept art, realistic, dark theme, low light,
|
Concept Art DarkSide is a merge of 3 models
0.30 - Concept Art Dark
0.35 - Concept Art Night (Unpublished)
0.35 - Concept Art Night Not Edited (Unpublished)
Recommended LoRA Strength: 0.7-1
For one reason or another, I don't publish some styles. I don't always manage to create a good style the first time, sometimes I have to merge several models to get the style I want.
kohya_ss does not allow to merge models with different Network Dimension (Rank). That's why I don't change the LoRA training settings.
