Hey piece of advice for LORAs like these, Don't leave stuff like angles, positions, framing tags in, (unless you want to try and attempt move the camera and framing or associate that prompt with that framing) as It'll confuse the AI. instead you'll want to leave in interchangeable things in the scenery like characters, character details, clothing, background, background objects, artstyle etc.

Make clothing tags as generic as you can unless its to re reinforce an idea like like a hyperspecific piece of clothing associated with that tag to recreate it. Your Dim and Alpha is too high for a 1 image dataset LORA as well tune it down to 8/8 instead of 32/32 as chews up space for just a simple concept. Either way good job. Its probably faster do just do things like these on a Control Net model but that's just me.