🎞 LoRA is trained purely on anime screencaps
🗝 Activation tags are in the right panel
📝 Note:
Her swimsuit has a high tendency for using a different style - I tried to get rid of bad images from the dataset, I tried to retrain her 4-5 times, either it's the ckpt issue I'm using or some combination of tags is poisoned on illu and there's night I can do about that - keep that in mind
Her suit is a bit unstable as there were barely 5 shots of it in Darkness, and I'm not using any images from the 1st season due to a different style
If the anime screencap style gets messed up, remove outfit trigger from the prompt.
📸 Always use ADetailer to get detailed eyes
Recommended positive prompt for adetailer:
<lora:ToLove-Ru_LalaSatalinDeviluke_IlluXL:0.9>, eye focus, purple eyes, no pupilsSame negative as for regular image gen.
🛠 Technical notes:
LoRA is compatible with various styles - Lower the weight & remove "anime screencap" from the prompt, if you are getting too strong anime style
Hi. res fix used on all sample images:
Upscaler: R-ESRGAN 4x+ Anime6B
Steps: 10-15
Denoising: 0.3
📬Leave a like, comment, review but first of all, ENJOY ;)