>>78576092A good rule of thumb for overbaked loras is always either
>too much details>forcefully changing your prompt into the lora (ie; changing hair colors, clothing etc)>making everything look deep fried A good lora should be flexible enough to make use of the zero-shot advantage that Stable Diffusion employ; If you can't prompt Kitanya wearing say, a white dress and a sun hat then your lora is shit. A bad lora will force her default clothing (and whatever was on your training data) and will end up deepfrying your results.
