I've found an issue saying I can train my character lora only on images, is that still correct? Is there any specific settings I have to set while training only on images? I'm planning to train two characters: first with a 60 image dataset and second with 120 images.