Trained on very small dataset 2 promo images + generated images
Notes : not my best work, it was mostly a proof of concept let me know what you think
PS : you can also use the tag "boots or Red boots" for full body photos
Trigger = dgbybr, Tags = 1girl, solo, brown hair, long hair, blue dress, black wrist cuffs, brown eyes,Description
Comments (2)
I think its worth exploring in-game screenshots (or screenshots of a ripped model) over synthetic data for this character. Pony and (so far) Illustrious have handled majority 3D and substantially 3D training data just fine for me when the data is tagged "3d" (and likely "low poly" in this case, though her model is pretty good for a DS game, especially with the ability to render the game at higher than intended resolution). When I see a model that suffers from having 3D data, it always seems the creator didn't tell the trainer it was 3D. There's also a few 2D CGs in game (some of which may need "chibi") from what I've seen of people playing it.
Hey nice i was wondering if someone made a lora model this game and now i see it, its great to use. Sadly that the game didnt recognize back then until decades later

