>>83811657>>83812357Alright, I did a few tests with it and it seems relatively usable, but there's a big issue with it forcing the model's art style and expression, the eyes being aimless in nearly all gens, and it leading to artifacts if you try to use other loras with it. It might not seem like an issue because if you're using a single artstyle lora you can crank it up to 1.5 weight and it'll "work", but the image is going to start melting and you'll have even bigger issues when pairing it with other loras.
I take it you're the one that made it (and the same guy that posted that zip with a bunch of geega images in lig and vtai)? If so, here's what I suggest in case you choose to do another iteration:
>Grab all art of her from danbooru https://danbooru.donmai.us/posts?tags=geega_%28vtuber%29+&z=5There's not a whole lot so it shouldn't take you that long, but there's way more art here than you had in your collection, a good portion of it being full body too which you lacked
>Remove most of the screencaps of her face from your datasetNot all of it, but the amount of them you have is excessive and is most likely responsible for why the lora is currently so stiff
Thank you for putting this together though, it's already more than I'm capable of doing in my machine kek
