Trained on AnyLora
Please note that with some enhancers(Detail Tweaker, e.t.c.) it is very easy to get better results than showcased, since i didn't use any to generate showcase images for this LoRA
To get the best results, please read the description below:
Recommended weight for V1: 0.6-0.8
Recommended weight for V2: 0.65-0.85
Activation word for V1: BoaHancock
Activation word for V1: BoaHancockV2
Use of "Detail Tweaker LoRA" is also recommended for good results
Upscale to 1.5
Using anime SD models is recommended, semi-realistic checkpoints also work quite well
I would appreciate any reviews, positive or negative!
I would very much appreciate you sharing your results ^_^
If you have any suggestions on what LoRA i should do next, please share them in the comments
Enjoy :)
Description
Version 1 - trained on 147 images
FAQ
Comments (11)
The output is decent some of the time,but a lot of the time, you'll get a giant snake/tail in the background. Perhaps you could crop the source images to just the character and retrain the Lora? Otherwise a pretty decent outcome.
I did kinda try getting the snakes out of the picture!
I always handpick my images and draw over them to cover unneeded elements. It was easy enough erasing the snake's head out from the images, but their bodies are usually so close to the character that it would just take a lot of time to clear them all without damaging the artwork :(
But i definitely will retrain the LoRA with better images, so stay tuned :)
Gotcha. Thanks for the response. I like your LoRA as is and if you make it even better, that would be awesome!
Thank you! I will try :)
Beautiful Lora, Tbh i think you'll agree with me that Boa Hancock without earring lora would be better? purely because the earrings are very prone to ruin the image?
Solid lora nonetheless, you're amazing as well :)
@PM786 Do you mean that the showcase images would be better without earrings? Or that a separate Boa LoRA without earrings would be better? There is no need to do a different LoRA for Boa without earrings, removing the 'earring' tag is enough to make it not generate any, as a last resort you can put a negative earring tag, if that's what you meant of course :) But generally yeah, it is easier to generate a good image without those earrings, because they are of a very particular shape!
@12user34kn276 Yes I meant that a new lora without earrings would be the best Hancock Lora in my opinion.
And I tried to put "earring" and "earrings" in negative prompt with weights and it 95% of the times ignore it. But that's totally upto you, I don't wanna trouble you needlessly, maybe you do V3 of this lora you can implement it, then it will be BEST Boa Hancock lora. Thanks anyways <3
@PM786 Thank you for the feedback! It worked for me for some reason, but maybe i haven't tested it enough :) A Version without earrings would be V3 already, and i don't think there is enough training data of Boa without earrings for such a LoRA unfortunately
@12user34kn276 can you kindly give me the negative prompt ? let me try it, maybe I am doing something wrong
@PM786 You’re not doing anything wrong! Sorry i was a bit unclear in my previous answer, i meant that i couldn’t get it it to generate without earrings 100% of the time either. But the usual formula for removing any unneeded element from a Lora would be lovering the weight of such LoRA, removing the positive prompts of the unwanted element(In this case removing earrings, jewelry, ornament and so on) and adding a couple negative prompts, ideally with bigger weights (earrings, earring, jewelry). Hope that helps :)
@12user34kn276 thank you so much <3
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.