Just a random fella doing random things in his spare time. Feel free to use the model or download it.
Any feedback is welcome!
The lora was trained using a trigger word:
triple-domain-expansion
Other useful prompts:
comic
english text
Use a resolution width > height.
Tools for regional prompts help.
Around 0.8 strenght is good.
Gloating hair in the negatives helps, as the dataset was a bit too repetitive due to the lack of images.
Text adherence is a toss of dice as usual.
Common Notes:
Trained on base Illustrious.
Strength between 0.6-1.0 usually does not give troubles.
It should work with other Loras without strange conflicts.
It should work on NAI too (mostly on the epsilon pred). If it gives some anatomy aberration, try without the main trigger word