This is the reason I started trying to make LORA's. The anime may not have been for everyone, but it's been my favorite since I first saw it.
It was trained on about 275 images taken directly from series stills focusing on many of the main characters and a few of the supporting characters also. With that large pool of images it has a habit of leaning towards the same style as the anime but with some work it's still quite versatile.
Sometimes needs helper prompts to correct hair color/length and eye color. Start without them and add if needed. Often needs (cat ears, fur) in negative prompt, most checkpoints think the Diclonious horns are ears.
The (Sweater) trigger is the one Mayu wears.
The (MaskedLucy) trigger does not work well. There were very few scenes to pull that look from, but I felt it necessary to at least try. Maybe you'll have better luck.
V9 - was the ninth time I attempted... it kinda worked and is workable... mostly.
V3 - As in the 3rd release, was trained on all the same images and captions as the previous version but using the methods I learned making 3 of my other LoRA's. It is both better and not better than it's predecessor. It works better for semi-realism but not as well for "less" semi-realism.
Description
This version now handles the (Diclonious horns) prompt better. There is still a tendency to generate cat ears, and with the addition of (horns) it will sometimes add other horns, thank you negative prompts. I've also downsized it considerably.
FAQ
Comments (14)
Really happy someone finally made an Elfen Lied model. I had trouble getting the headband and horns with Nana. It often made the horns cat ears also. Maybe you could train one on only Lucy or Nana to go with this.
most models I use this on has that issue. I found that I often have to adjust the negative prompt to account for it. As far as the headband, I didn't think to train it specially for that so inpainting might be needed there. I'll try tomorrow to see if I have it in the dataset by accident.
also of note, do you have V6 or V9? there's not much difference with headband related training but it focused more on getting the (Diclonious horns) trigger working better.
@Avenka I used V6 for the ones I posted. I downloaded V9 and actually got some good results using it at .5 while still using the V6 lora.
@ai_degenx I finally remembered to look, try (bandana), it's not "trained" into it but it gets you a starting point.
@Avenka Thanks, I will give that a try.
Girlpack (with guy) YAY!!!
Will I be able to create them in the anime style, or will they be this detailed? I am hoping to make them look how they originally did.
This was created to have flexibility. That being said, most of my LORAs are still capable of what you're looking for. Often heavier weights and the proper model ckpt gets you there. I'm not at my computer to test but try things like increasing the LORA weight from 1 to 1.1 and putting heavier weights on the character ie (Nyuu:1.3) I would also suggest more anime leaning ckpts like 2dn_2dn.
@Avenka I've never heard of 2dn_2dn! I use anything v5. I wonder if that'll do... I'll try as soon as I can <3
@xSatouMatsuzakax I made a post with 2 examples of what can be done with some quick prompt tweaking. One with Anythingv5 and one with 2dn. These are far from refined, short prompts and no impainting but I think they give a good idea of how close you can get.
Hope the examples help
@Avenka I logged on and tried to use it. Even with inpainting it refuses to create the horns, and the style is wrong. I think I'll probably delete this lora for awhile. I don't want to have to get an entirely new checkpoint just for one lora. Thank you anyway.
no problem. I'm sure I am in the minority of people who want some flexibility in a Lora. Most people want a 1to1 copy while I want to be able to see what it should be but not get the exact same thing. No hard feelings.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.