trained a Yagami Light LoRA (0.7 weight, recommended with Hirefix) , lora trained on civitai ,I generally avoid training here (mostly for Stable Diffusion), but Colab problems necessitated it. The results are mostly decent but imperfect.  Consider including ,multiple  people, two boys for negative prompts for better outcomes. 

recomended model are anylora, cat, and cartunafied v3

try my lora in https://tensor.art/models/805004380530277859