So we need to create more LoRA models using Pony of specific characters.
To do this, describe the character intimately instead of using the coliquol terms like instead of Bee Girl, you do Bee Girl, yellow skin, black and yellow stripes, bee wings, mandibles, thorax, 2 antenna, etc.. and sift through the images, making sure to like and dislike them and add to the negative prompt whenever something isn't right, and you will acquire enough images to train a model so you never have to do that again.
Alot of the time, if you do purely by name you will get a 50% approximation, unless you use an already trained model or you describe them intently.
Additionally, if you do not have the funds or the time to train a model, post the images. Somebody else can use them as training.
Eventually all of these characters will be crunched down into "packs" where the few commonalities become a "style". Kind of like how texture packs work. Currently there are quite a few of these "packs" but they are built mostly on cliche or normal scenarios, so mostly NSFW or absurd imagery. That is why it is important to train your own models for your own generation, ultimately the only way you will get THAT image, is to make it urself, with models.