Chel from Road to El Dorado.
1000 images as dataset, no AI generated images were used.
THERE WERE SOME ERRORS IN THE TRAINING DATA SO I RETRAINED AND UPLOADED IT AGAIN. PLEASE DOWNLOAD THE NEW LORA.
Description
I used a vaster dataset (250 images), more diverse and better cropped.
Overall the results seem to be better but only with WAI 13, as 10 and 14 provide more anime-like images of Chel, which we dont' want.
Provide feedback please.
Trigger word: ch3ldy
latina girl, dark-skinned female, bangs, long hair, black hair, brown eyes, wide hips, sexy body, fit body, thick lips
FAQ
Comments (13)
It's not working
What do you mean?
@thegreatensing336 It's not generating her accuratly
@Steeltron2000 I did my best to create a Lora that generates the most accurate version of Chel.
@thegreatensing336 Ok
@thegreatensing336 It works very well for me, I hope you continue making more Loras! They are the best I have seen
@MegaEevees thank you very much :)
@thegreatensing336 I see why Steeltron2000 was complaining. Your loras are amazing and super accurate however there is something odd with your loras. They dont work in a1111, I can make them work in Comfy. Any idea why? I rather use a1111
@whiteowl I have no idea why comfyui and webui work slightly differently, I should do some research. If you discover something, I would love to know.
@thegreatensing336 Sure, lets try to find out how to fix it so everybody can enjoy your loras no matter what they use. Your loras are really really good, congratz.
@whiteowl thank you very much :) for reference, I use civitai to generate my Loras
@thegreatensing336 I must really like your loras to spend 2 days investigating hahahaha. I found the problem, is the naming. You use double underscore "__" in the lora name ss_output_name and modelspec.title. When a1111 sees the __ in the promt and you have the wildcard extension in a1111 the system thinks is a wildcard and not a lora. So the lora is never applyed as a lora. All you have to do is update the metadata to use only 1 underscore "_" instead of 2 like Jasmine_Aladdin_Illustrious instead of Jasmine__Aladdin__Illustrious. Im doing it manually but maybe you can reupload all your loras with this name update so no one will have this issue when they download your loras. What do you think? (ofc this is for those loras with __, Chel lora doesnt seem to be that case)
@whiteowl wow that's brilliant. Thank you very much, you did great :) I will start doing that at some point
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.