Kindly, PLEASE READ THE DESCRIPTION BEFORE USING THIS LORA (VERY IMPORTANT ⚠️)
IMPORTANT NOTE ⚠️ :
Please note that this LoRA is specifically designed for male characters. It has not been trained to generate female characters, unless combined with other LoRAs that are specifically trained for that purpose.
Use tsubaki or coco-illustrious-XL(https://pixai.art/en/model/1825497827039510248-coco-Illustrious-NoobXL-Style) model for the best results with this lora.
make sure to put the weight of this lora between 0.7-1 (if you're using it alone)
If you want to use it as secondary style-loras then put the weight around 03-05
Optional negative prompt : wet skin, ambient lighting, ugly, unattractive, oblong face, sharp teeth, fangs.
Difference between v1 and v2 (noob-ai-xl and tsubaki)
V1 noob-ai and tsubaki pros & cons :
- Semi-realistic.
- Give you more accurate masculine and handsome results
- Struggle with alot of perspective, pose, and composition too ig?
Best settings : 20-25sampling and 3.0-6.0 cgf scales (higher is ok)
Best settings : 25sampling, 7.0 cgf
☁️☁️☁️☁️☁️☁️☁️☁️☁️☁️☁️☁️
V2 noob-ai pros & cons :
- More into anime/cartoon style rather than semi-realistic.
- This time i didn't overfeed the ai with similar image so hopefully it didn't struggle with various poses and angle (i hope so)
- More into youthful yet masculine appearance.
V2 tsubaki pros & cons (new) :
- detailed lighting & High contrast and quality.
- Compatible with both, serin and tsubaki
Best settings for serin model : make sure to put 15-20 sampling and 5.0-7.0 cgf scales, otherwise you'll either get an extremely high contrast results or disoriented / blurry shading and rendering
☁️☁️☁️☁️☁️☁️
V3 noob pros (new) :
- Still semi realistic.
- High contrast and detailed lighting
- Can create both, feminine and masculine characters
After experimenting with this version for a long times, I've finally come out with some tips.
The best settings for coco illustrious V3 model would be 15-18 samplings and 3.0 cfg scales (if you're using it alongside hypersdxl or any model enhancer loras)
But if you're using it alone without any "Hyper-sdxl" or "model enhancer lora's", the best settings i would recommend is 20-25samplings and 2.0-6.0 cfg
Sincerely, Thanks for reading the description and enjoy!
