@atomicpoet I tried to get Stable Diffusion to generate a custom PHP “elePHPant” (see here for examples of what I mean: https://elephpant.me), but it kept failing because it hadn’t been trained on anything that looked like that.
So, I had to train a LoRA on a bunch of photos of the plush toys. Even still, it’s limited in its ability to generate images of them from certain perspectives and angles because I don’t have enough training data to show the toys from all possible angles.