Heard someone wanted a Starbucks Siren LoRA, so here it is.

Kinda surprised that there isn't one already.

OB (OverBaked versions):

It's a simple dataset of 40 images of the siren and Wendy, and 1 image of Zelda, for fun.

My hope of including Wendy and Zelda is that it won't make everything too greyscale, so that it's more flexible. But test it for yourself and see how it works. Be advised, the OB versions were purposely overtrained to see what would happen. OB Pony should work fine at a weight of 1, but OB SD15 should be set to no more than 0.7.

Revised SD15/Pony:

Added 9 more images of Zelda, plus a couple images of Wendy, bringing the total tally to 52. Reduced the amount of repeats to something reasonable, and revised the prompts for each image, so it should work better. I will test them, but for now, assume that a weight of 1 should be fine for both SD15 and Pony.

Tips and notes:

Trigger word for the Starbucks Mascot is "Starbucks Siren" and/or "STB-chan", STB-chan makes her more stylized, so if you want a realistic approach, you can leave it out. There are different ways to depict the Starbucks Siren: monochrome with streaked hair, grey skin, black and white hair; green hair--and be sure to leave out monochrome/greyscale prompts to give her color; pink and green streaked hair and/or "STB-Chan \(Sakura\)", which will give her an even more colorful look.

Especially with the Pony models, Wendy can't simply be Wendy; I've tried. The trigger for the Wendy's mascot should be "Wendy \(Wendy's\)", otherwise things tend to get inconsistent. For example, I think I got Wendy from Fairytail, so be advised.

Zelda is simply Zelda. Theoretically, it should only depict the BoTW version of her, but if you find that it changes to something like ultimate smash Zelda, then that would be cool, too XD.