Trained exclusively with anime frames to avoid art style mismatches.

-Trigger: 'Haruka_ORAS'

-Weight: Between 0.7-0.9 (lower it further for more varied clothing and poses. You may have to add some stuff back if it disappears, as 'red hair bow', 'blue eyes', 'yellow fanny bag', 'bike short under shorts'…)

-Considerations:

1. The red shirt may show up shorter than usual, exposing her midriff. It looks cool, but if you want to prevent this, add 'midriff' and 'navel' as negative prompts.

2. The reference frames are taken from a low-quality trailer, so always try to include '(blurry, unfocused)' as negative prompts to compensate.

The LoRA is trained to include her usual clothes, so if you want to get rid of them, write them as negative prompts and lower a bit the weight.

I tested with many checkpoints and random stuff like 'wedding dress', 'playing soccer' or 'badass punk girl' and they worked fine… so I guess that means the model is flexible? dunno