(Do not Re-upload this To PixAI!!).
Want to Support for Future development? Go HERE.
Hello guys, what's up.
I will share with you my main checkpoint in recent months for creating desktop wallpapers.
In this build, I create this checkpoint mainly with more artistic style output, rather than Consistent stylish AI IMG build.
For the base Model I use (IL 2.0), NoobAIEPS1.1+_((Anzhc)noobAI clip swap), (plants mix (coconut))+ train Over 1k IMG (mixed Tags X Caption)+250IMG for style
For sampler, scheduler, steps, and Resolution
I recommended using
UNI PC CFG 4.0 SGM Uniform 20-35 Steps. Clipskip 2 Portrait 1024 x 1536
But I mainly using wide Resolution
Or use your favorites settingMain Output (2.5D) with slightly paint brush stroke
(Char Mix Male and female(male=Korean face Dominant))
Main prompting Style:Tags
Main Problem Some Anime character really hard to blend into certain style if you using long prompt.
Not very good using NLP since this checkpoint isn't DIT ModelFor dataset Cut-off, I believe those Noob AI clip around DES 2024,
I don't really bother with recent character data since the main purpose of this model to create artistic output, yeah maybe I added some HSR character, and other, but it's only my for character that I like. For NSFW, I don't really know the checkpoint can handle it very well, so feel free to test them out at your own risk.
I also Recommended mixing the output with some artist tags that you can put in front of your main prompt
you can found over 100 Testable Artist mixing over --->Here⇽--
For Prompting You can use mix both Natural language and Tags for the best output
(you can see on the Sample IMG).
Do not Use Score
Use Simple negative as posible then add more negatve keyword if you feel something offFor License
You cannot Sell directly the generated images for commercial, If not Fully Edited or paint over it.
Use it for Fun Aka Hobbyist
IF you have a problem or suggestion for the checkpoint, let me know in the comments.
# Stop doing Stupid early access. Just download it directly
# Checkpoint model Should be open source and downloadable because those are using copyright IMG material!!!!!!
#OMG this checkpoint suck. (simple, give me some advice for future version then delete them on your hard drive)
Description
If you never try, then you should try
High-res output Up to 2K res #liar, this one just marketing. (#use your favorites personal resolution)
Train over mix Tags and caption using 1536 and 2k IMG / bucket
Artistic Output :p / depend on what you want to achieve, #Simple prompt always better
can be mixed with other artist style, #Maybe just test them
Lora compatible, Maybe
Direct Download. Yes
I don't recommend using any Lora booster like DMD,sdxl lighting or any similar that can make your output faster (it can break model output)
Suck at NLP because this one is not DIT, #use Neta Lumina if you really want using NLP















