It just works.
Generate your character surrounded by unsettling eyes in dark portals.
Fun Fact: There's already a large sample size to draw from, and anime models already have this tag built into them. However, the concept is so diffused that nothing of it can be reliably generated if at all. This LoRA concentrates all of that, specifically curated from a dataset of over 8000 pictures, where I handpicked only 100 aesthetic high-impact, dynamic shots for the best possible generations.
Generations are largely random, but I get the feeling you can use inpaint with great results.
Works fine at a weight of 1. Adjust as needed.
Play with the weight of both the tag and the LoRA and see what happens.