Helluva LoRA, Helluva Boss Multi-Character LoRA by ArturroWolff

https://civitai.com/models/239248 Please use e621 tags as trigger words for characters, i.e.: "stolas (helluva boss)" or "moxxie (helluva boss) This upload includes various epochs of the same LoRA, the creator has noted the instability, so use with the knowledge that results may not be optimal. Characters/styles included: Stolas, Octavia, Stella, Blitzo, Moxxie, Millie, Loona, Fizzarolli, Asmodeus, Queen Bee, Mammon, Arturo Wolff, Stelloxxie, Stovia, TestShip Clip Skip 1 Author notes, ArturroWolff: Welcome to the ultimate Helluva-verse and Hazbin-verse LoRA!

Not really.

This LoRA is the result of my mental health decline and living proof that quantity doesn't mean quality, and that efford goes on these things for a reason. A self-imposed challenge to see how far I could push common sense out of the picture to create something that shouldn't work yet it does just to spite god and his creation.

This LoRA was trained on 26,962 images scrapped directly from e621 and compiled together on a 1 repeat folder and a whopping 2 training batch size. 13,481 steps per epoch of pure chaos every 10 hours.

Can it pull it off?

Not really

Since this was done out of spite against anything holy, the training data had nothing but the original e621 tags formatted into them. Here is an example prompt:

octavia (helluva boss), owl, owl demon, bird, avian, female, solo, formal clothing, detailed background <loragoeshere>

You can alter this to fit your model's tags, but I recommend keeping the character tag the same "octavia (helluva boss)" and the meta tags like "helluva boss" and species tags like "owl", "owl demon", etc.

If you wonder which tags does a character have on e621 simply look it up, copy paste it, and have fun.

"I have X issue"

I bet you do

Since this was a meme proyect, I won't provide much assitance or help, if anything feel free to do with it whatever you wish, all perms are on for you to go crazy.

Each version is an epoch so you can choose whichever you liked the most based on the XYZs provided in the gallery for version1.

Only tip I can give you is: Skip 1, fluffyrock w/vpred. Those were the settings this LoRA was trained on.

What now?