Devilman OVAs (1987-1990) style -- LoCon for Pony

A huge Pony style LoCon for the classic Devilman OVAs -- "Devilman: The Birth" (1987) and "Devilman: The Demon Bird" (1990). These OVAs are slightly different stylistically, but they were both produced by Oh! Productions and Kodansha, and the final LoCon is pretty good for my tastes.

A quick plug: I have other classic anime style LoRAs/LoCons. Check out https://civitai.com/user/SweetRammaJamma!

_

UPDATE, MAY 19TH, 2024:

I don't recommend that you download these, but for the record, I've uploaded all epochs of this LoCon to a huggingface repository here: https://huggingface.co/SweetRammaJamma/DZVZVZ/blob/main/README.md

I happened to use the same dataset as the LoCon to test "plain LoRA" and LoHa versions of the same net filesize (~900 megabytes), and those versions are also in the repository. For what it's worth, I didn't detect any huge differences between those three LoRA types. My guess is that scale_weight_norms=2 and dimension=64+ resulted in the LoRA, LoCon, and LoHa being so similar, and more reasonable settings would reveal more pronounced differences between the three types.

The net result of my testing was simply assurance that LoCons work fine for my purposes! I hope to someday learn to exploit the differences between the various LoRA types, but for now, I'm just glad to have some settings that get fairly consistent results when training Pony LoCons on a variety of datasets.

__

SOME INFORMATION SPECIFIC TO THIS PONY LOCON:

Primary trigger word is DZVZVZ (DEVilman OVA, with each vowel replaced by a Z).

FOR PONY CHECKPOINTS, TRY: 1) ColorburstXL, 2) LuminaverseXL, 3) Reponygine, and 4) "Prototype" (see the "GENERAL INFORMATION" below for more info, as well as "RECOMMENDED RESOURCES" at the bottom). Keep in mind that there's non-Pony versions of some of those checkpoints, so be sure to get the Pony ones. Also, most of my testing is with "primarytriggerword, score_9, score_8_up, score_7_up, score_6_up, source_anime" at the beginning of the positive prompt, so try that first.

Characters were tagged with the following trigger words. Disclaimer: Generally speaking, many of the character tags work quite well, BUT, "specific character likenesses" were NOT my actual goal. Please understand that main characters may be overfit because of the sheer number of images. Less common characters are likely to be either overfit or underfit (because they have so few images, and/or their images all share very repetitive features like identical clothing, identical backgrounds, etc).

MCDevil (for Devilman)
MCSirene ... Due to the source material, some purple tinge/glow might appear at the edges of the image, or at the tips of Sirene's "head wings." I tried to address this through clever cropping AND (when appropriate) the tag "glowing purple border." So if you do get that purple around the edges, you can TRY "glowing purple border" in the negative prompt.
MCMiki (she's basically the only HUMAN female character in the OVAs, so consider using MCMiki if you're getting "masculine 1girls")
MCNakira (this is "nice Akira," from the first OVA, before Akira becomes Devilman and becomes much more masculine -- tagging these "two Akiras" separately was especially important for the differences in their eyes)
MCBakira (like "badass Akira" -- from the second OVA, when Akira becomes more masculine and serious-looking)
MCRyo
MCGangster (used to indiscriminately tag every gang member from the first OVA -- the results of using "MCGangster" will be chaotic, because the tag doesn't refer to any specific character)
MCGhelmer (Ghelmer doesn't have many "full body" images of him, or really many good pics of him at all, so be prepared for "abstract interpretations" of Ghelmer)
MCJinmen (similar issue to Ghelmer -- be prepared for things to be either overfit/underfit)
MCProf (for Ryo's adoptive father)
MCKaim (the same issues as Ghelmer and Jinmen, but even worse -- things will DEFINITELY be either overfit or underfit, here!)

Also, I tried to absorb scenes of Jinmen's "shell faces" into a single tag -- so try "horror \(theme\)" if you want to experiment with that. But those "shell faces" are pretty repetitive, so prepare for "stubborn" results.

Other than Sirene and Devilman, every monster was tagged with "monster." In the case of named monsters, this went "MCname, monster." But many, many other monsters were tagged with "monster" as well. And when there were multiple monsters, I tagged them with "multiple monsters." So, monsters might be fun to play with!

_

And yes, this Pony LoCon is slightly LARGER than my other (annoyingly huge) Pony LoCons. The training settings were slightly different than some of my later LoCons. But I really like this LoCon as-is, so, I'm not going to re-train it with new "standard settings."

_______

SOME GENERAL INFORMATION ABOUT ALL OF MY LORAS/LOCONS:

To generate the sample images, I don't use outside LoRAs. As in, I only use the checkpoints plus my own LoRA/LoCon. I don't use embeddings, either. So the sample images are very "pure/simple."

I use ComfyUI only, so my image generation methods might be different from what's typical with Automatic1111 or CivitAI's online generator. Notably, I can't inpaint in the STRAIGHTFORWARD/POWERFUL/EASY Automatic1111 way! If things get too weird, I usually just "roll the dice" again! But chances are good that inpainting will be an asset to you, especially for things like getting eyes that are 100% stylistically correct.

99% of my testing is with Euler A, at 25-30 steps, CFG 6-7, with the LoRA/LoCon at strength 1.0, and with the trigger word at the very start of the positive prompt. I test both latent upscaling (1.25x to 1.5x, 0.60-0.65 denoise, ~12 steps) and "regular" upscaling. I've tested face detailers, too. All of it works appropriately, but some typical "light troubleshooting" will be required if you're going for a specific output.

I do upscale my sample images quite a bit. Usually, it's latent upscaling first, then an additional pass with "regular" upscaling at about 1.5x. I usually resize with boring old Lanczos, and not "model upscalers" (such as AnimeSharp4X), because Lanczos resizing ends up more faithful to the original anime's style -- but really, I do whatever works.

In regards to upscaling and YOU: Don't be surprised if something like my Guyver OVA LoCon gives you gritty/grainy initial results. Experimentation is required. And it varies from LoRA to LoRA -- for example, FLCL is a nice-looking and somewhat modern anime, so it doesn't have the same problems as the Guyver OVAs.

Where you put the trigger word does matter, and (the trigger word's weight:1.2) does affect things. So to decrease the "faithfulness/literalness" of the style, put the trigger word all the way at the end, and (decrease its weight:0.5). Of course, you can play with the weight of <the actual LoRA/LoCon:0.8>, but I tried to ensure that things work reasonably at 100% strength. (If my LoRAs/LoCons are totally unusable at 1.0 strength, I simply don't release them! I find another way to re-train them!)

For both Pony and Animagine, I try "simple background, gradient background" in the negative prompt. This does seem to help with the typical boring backgrounds of SDXL anime models, and it might help more than usual in cases where a LoRA/LoCon is based on an anime that has many boring/simple backgrounds (FLCL comes to mind).

In general, my prompts are uncreative. I just use wildcards -- which is partly why, if you look at the image generation data, you'll see some "dead weight" words that don't seem to be doing anything. Also, I'm very lazy, and I don't even bother removing tags like "worst quality, low quality" when switching from AnimagineXL to Pony. So, I gave you my actual prompts in the image generation data... but don't assume that my prompts are worth imitating.


___


GENERAL INFORMATION FOR MY PONY LOCONS:

My Pony testing was all done with pretty standard quality tags in the positive prompt, in the form of "primary trigger word, score_9, score_8_up, score_7_up, score_6_up, source_anime," so I can guarantee that those work, but feel free to experiment!

For checkpoints to combine with this Pony LoCon at 1.0 strength, I recommend starting with the following (which are in the "Recommended Resources" at the bottom of the page):

1) ColorburstXL (Pony), because it produces "faithful style" while still being more consistent and aesthetically pleasing than the base Pony model. Then, maybe try...
2) Luminaverse (Pony) will get you something more "modern" while still being quite faithful to the LoCon.
3) Reponygine (Pony) will get you something quite "modern and pleasant," while still retaining good core style features (like eyes and hair details) and character likenesses.
4) "Prototype" (Pony). I don't quite know how to describe the look yet, but it works well.
5) AnimeConfettiComrade (Pony), IF you want a MORE "literal" interpretation of the LoCon, while still being more consistent than the base Pony model.

You CAN use this Pony LoCon with the base Pony model, but the style might be overpowering, and you may have to lower the LoCon's strength a fair bit below 1.0, and/or move the trigger word to the end of the positive prompt, and/or reduce the strength of (the trigger word:0.1) -- which are all the same ways that you'd adjust the strength with other checkpoints, of course. Adjust to your taste, always :) And, there are many more Pony-based checkpoints that I haven't even tried, so if you find one that works well with this LoCon, please let us all know!

_

Yes, my filenames are long. It's because I personally use LoRA filenames to quickly compare different versions of my LoRAs while testing things. I recommend renaming the file to your tastes! Though it would make sense to leave the primary trigger word in the filename, right? (It's up to you!)

Yes, my Pony LoCons are extraordinarily large. But I made them for myself, and, to quote my genius friend, "it works for me and what I do." If you try these Pony LoCons, and you've made some SDXL style LoRAs/LoCons that you like, I would be happy to discuss training settings with you.

__

No warranty, express or implied.

I place no special restrictions on what you may do with these .safetensors files. Likewise, I am not responsible for anything that you do with them.