I'm only gonna say this for you to improve alright?
Bad dataset = bad LoRA.

Your dataset contains only 53 real images, which are for most either noisy, too small and for most of them contain a watermark. Training LoRAs don't work like that sir, with all due respect 🪵🐸🪵

Hey I tried this, thinking it would have a effect with some other gape/anal lora. But somehow it doesn't really do that much? :0

I tried using your prompt from your images, but this is using a reference image. And I was hoping for a model that works by itself. Do you have any examples of standalone generations? Your dataset is pretty large 460MB, which is pretty big for something of which I don't know if it works. Can you make some examples with your lora alone?