The model I think is fairly self explanatory.

This ones a bit of an experiment on the Omnibus variants I do, by classifying each unique doujin in the lora with it's own unique tag in addition to an Omnibus tag, seems to have worked perfectly. Each unique character in each unique doujin is able to be pulled out of each lora no problem. For the Isekai doujin to differentiate between characters I'd suggest using descriptive tags such as cat ears, wolf tail, demon horns, demon tail and such. Largely this experiment was a smashing success.

Previews made with novaFurryXL_illustriousV20.