THIS IS MY MOST AMBITIOUS MODEL. ALL THE WORK HAS LEFT ME EXHAUSTED, SO I HOPE YOU APPRECIATE IT AS MUCH AS I DO.
If you liked it, don't forget to give me some (buzz)
I RECOMMEND READING THE MODEL INSTRUCTIONS
-The model was trained with 1500 manually edited, classified, and labeled images.
-The model offers great flexibility and performs extremely well with danbooru thanks to manual labeling.
- All images in the gallery were generated solely with prompts; addetailer was used only for the face.
PLEASE DO NOT POST NSFW CONTENT HERE
Lilandy-Ultimate-Style-(Anima)
-now works with almost any checkpoint
RECOMMENDED WEIGHT: 1
Steps: 20-30
CFG Scale: 4-5
addetailer (It's not necessary if you don't want it; Anima does a great job with faces)
Sampling Method: ER SDE
Schedule Type: Beta
Recommended
Keyword:
-lilandystyle, 1girl, solo,medium breasts
-lilandystyle, 1girl, solo
Negative-(Recommended)
-worst quality, low quality, score_1, score_2, score_3, artist name,jpeg artifacts, bad anatomy, bad hands, bad teeth,realistic teeth,3d,
-scary teeth, sharp teeth, (optional)
Lilandy-Ultimate-Style-(illustrious)
-now works with almost any checkpoint
RECOMMENDED WEIGHT: (0.8)-(0.9)-(1)
Steps: 20-30
CFG Scale: 7-9
addetailer
Recommended
Keyword:
-lilandystyle, 1girl, solo,medium breasts
-lilandystyle, 1girl, solo
Negative-(Recommended)
-low quality, lowres, worst quality, watermark, bad anatomy, jpeg artifacts,scan artifacts, bad hands,bad eyes,deformed pupils
-face tattoo (optional)
Description
FAQ
Comments (11)
You've finally finished this amazing Lora. Thanks for the awesome work!
It was a lot of work, but it was worth it. :)
More than 100%. ;)
With a that high CFG value don't you generate over-burnt pics?
Also regarding recommended tags, some aren't danbooru ones. Sure ILXL models have somewhat of a natural language understanding through OpenCLIP ViT-G/14 and CLIP ViT-L/14 but this is still miles away from models that have modern clips. So using natural language with Illustrious is mostly like flipping a coin and hoping it lands on the side you bet on: inaccurate/reliable. So unless you tell me you've trained your LoRA on tags such as "bad eyes" or "deformed pupils", i would be rather skeptical about their effectiveness.
All in all i d be curious to know which tool you 're using to generate pics.
Hi, I'm glad you like my model. I recommend using cfg in version 7. Regarding the labeling, it was done manually, although it's not perfect and there may be some variations in character creation. As for character generation, I use NeoForge based on Automatic111.
Awesome style!
Hell yeah! Your styles are top tier!
So you didn't tag specific characters with this? Darn.
It works well with Booru and fits the style well; in the future, the character tagging could be improved, but I still have many things to do.
Very good lora. Solid in the way the style is represented, clean lines, few errors in line or anatomy. Might be your best one yet.
I know what you are.