>>4300
I think it would be fine. But even if it doesn't work no harm in trying. Dinosaurs & dragons are already trained in the base model so I don't think we would need a lot of monster specific training data. As long as we have enough for the biting/eating it should just translate to any creature. To answer the question since we would be training the concept of biting/eating as long as there is sufficient data in the base model it should be feasible to mix monster types easily. Which I believe there is. If it made properly it should mix well with other monster lora's like pic releated, so you can always just pull monster data from other Lora's

For the impalement Lora I would need at least 20 images with people being impaled on a stick of any kind. Shouldn't matter. If I recall correctly I didn't even have 20 which is why I think it failed. I wasn't very good at collecting images back them. Although I'm still not tbh but especially so back then, I had no idea how to navigate gelbooru/pixiv properly.