Version 1.1 is now released. The Tag "One piece" is now integrated in the model. Because of that Gloriosa look much better now!

Hello everyone,

before i start:

Happy new year!!!🥳🎉🥂

Today I’m posting a requested model of Gloriosa from the One Piece anime.

At first, I wasn’t happy with this LoRA — it was hard to create, because the character only has a very short appearance (their younger form!) in episode 1130.
So I made a hybrid dataset that included:

• anime screenshots
• fanart
• ControlNet-colorized images

ControlNet was a real pain because I couldn’t get the style right at first. Then I thought:
“Wait — there must be some One Piece style LoRAs out there.”
After finding one (see below at Suggested Resources), I was finally able to generate some good results:

This one is a colored version of her introduction in chapter 1156.

Since I used so many different sources, the styles kind of blended together, so the results were pretty average at first.

Luckily, there’s a tag that improves this LoRA a lot: “one piece.”
I didn’t bake it directly into the LoRA (maybe I’ll add it in a version 1.1), but I highly recommend using it when prompting.

I did some images using the same seed for comparison. Check out the example pics.

Note: If you don’t want the cape, add “cape, epaulettes” to the negative prompt!

Prompt: look at "Trigger Words" on the right side 👉

Changelog: look at "About this Version" on the right side 👉

My Settings (Forge Classic)*:

*note, the example images only use high-res fix.

Checkpoint: waiNSFWIllustrious_v160

Sampler: Euler a

Scheduler: Automatic

Samplings Steps: 30

Hires.fix: R-ERSGAN 4x+Anime6B, 20 Hires steps, Denoising Strength: 0.5

Extension:

Adetailer (highly recommended)