/vt/ - Virtual Youtubers

As long as she's dressed the same (robot body) and looks the same (green/blue eyes, facial structure) there's no need for 2 loras. And even with different outfits, multiple concepts can be baked in a single lora triggered by a single word, like rushianecro, rushiapink, rushiaschool to trigger 3 of her outfits if the lora was trained with all 3.
Toon or not toon depends on the model that is used to generate the image, not on the training pics used. Same way I can make realistic images with vtuber loras that were only trained on 2d images.
Just don't expect 100% fidelity and flexibility at the same time. It's a tradeoff