Eh, to be honest I doubt you'd get it to do much without all the parentheses to alter the weights of words, especially if you're aiming to change outfits. The base Lora suffers from an insufficient amount of images for training (Tila doesn't actually have that much official art), so I suspect that's part of what causes the Lora to be so unwieldy and to default to certain outfits/poses more strongly. With enough tinkering around you can usually get it to do what you're wanting, but it's not very straightforward and takes a lot of trial and error. Someday I'll probably try to remake it with an expanded data-set including additional ai-generated images to give it more flexibility, but there's other things I'm working on for now.