Shin Sekai Yori style -- LoCon for Pony版本ShinSekaiYori-Pony-V1 (ID: 515468) 综合资源合集综合资源合集

A huge Pony style LoCon for the 2012 Shin Sekai Yori anime.

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

_____

SOME INFORMATION SPECIFIC TO THIS LOCON:

Primary trigger word is SHSZYZ (which is SHin SEkai YOri, with each vowel replaced by a Z)

All sample images were generated with ColorburstXL. I know that there are other good checkpoints, though -- please experiment!

Characters, creatures, and significant features were tagged with the following trigger words ... followed by additional common tags for that character. General disclaimer: Most of the more common character tags work quite well, BUT, "specific character likenesses" were NOT my actual goal. Main characters may be overfit because of the sheer number of images. Less common characters are likely to be either overfit or underfit (either because they have so few images, or their images all share very repetitive features like identical clothing, identical backgrounds, etc).

MCSaki ... (brown hair, light brown eyes)
MCSatoru ... (dark blonde hair, violet eyes)
MCMaria ... (magenta hair, light blue eyes, blunt bangs, hime cut)
rodent man ... (long teeth, red sclera ... or blunt fangs, yellow sclera)
multiple rodent men ... (army, military)
MCSquealer, rodent man ... (long teeth, orange irises, black sclera, orange pupils, bald, facial mark, forehead tattoo)
MCShun ... (dark grey hair, dark green eyes)
MCTomiko ... (dark purple eyes, grey hair, medium hair, wavy hair)
MCMamoru ... (yellow eyes, brown hair)
MCKiroumaru ... (green sclera, sharp teeth, traditional clothing, light brown vest)
MCShisei ... (eye mask, dark grey mask, yellow mask, double pupils)
MCMizuho ... (dark yellow eyes, brown hair, single braid, hair over shoulder)
MCFujnek, pokemon \(creature\) ... <-- for the "fujouneko" or "impure cats" ... (magenta sclera, horizontal slit pupils, yellow pupils)
MCInui ... (dark grey hair, grey eyes, traditional clothing, light brown vest, red sash)
MCFinoshiro, multicolored glow ... <-- for the first False Minoshiro
MCMinishiro, stuffed animal ... <-- for the second False Minoshiro
messy MCMaria ... <-- for "the Messiah" ... (magenta hair, turquoise eyes)
rodent queen ... (red irises, black sclera, red pupils, blunt fangs)
MCKoufuu ... (bald, facial hair, beard, mustache, prayer beads)
MCMonk ... <-- for most of the monk characters ... (bald, black mouth mask)
MCTakashi ... (brown eyes, brown hair, light brown hair)
MCSubaru, mutant dog ... (sharp teeth, tongue, saliva)

Note that some "character tags" are really pairs of tags. For example:

1) Squealer was always tagged with BOTH "MCSquealer" AND "rodent man" (PLUS "orange irises, black sclera, orange pupils" when his eyes were showing, PLUS "long teeth" when his rodent teeth were showing). This seems to have successfully improved "feature separation" on both Squealer and the generic rodent men. Combined with the "rodent queen" (whose eyes were always tagged with "red irises, black sclera, red pupils," and whose teeth were always tagged with "blunt fangs"), this seems to be a big part of why Squealer's eyes can be fairly easily applied to non-Squealer image generations.

2) The "fujouneko" (or "impure cats") were always tagged with BOTH "MCFujnek" AND "pokemon (creature)." This solved an issue I was having in early versions of the LoCon, where using single tags like "MCFujouneko" OR "monster cat" OR "mutant feline" (etc etc) seemed to cause the fujouneko to pop up whenever I prompted for just "catgirl."

There's definitely still some kinks that I could work out with some tags. For example, I considered changing "rodent queen" to "rodent kaijuu" (or similar), because although "rodent queen" currently does give interesting/funny results, it tends to put a fairly standard "queen's crown" in the image! And, I didn't test it, but "MCSubaru, mutant dog" could conceivably cause dogs (and Subaru automobiles?!) to pop up in images, similar to the issues I had when I was using tags like "MCFujouneko, monster cat." But, since I successfully debugged some of the bigger issues, I decided to not sweat the small stuff, and upload this current version.

Using just one tag from a tag pairing could yield interesting results. For example, "MCFujouneko" might get you some fujouneko features on non-fujouneko characters. Increasing the weight is probably necessary, like (MCFujouneko:1.4).

Some other tags:

hockey mask, white mask was used to tag Shun's mask when... you know.
organs, horror \(theme\), pink theme was used to tag... that torture scene... near the end of the series. Be careful with that.
monster, red background was used exclusively for Saki's strange vision with the really, really weird creatures that don't appear anywhere else in the series.
surreal was a catch-all tag for several unusual scenes.

And yes, a fun thing is that you can put Squealer's eyes onto just about anything by using his "orange irises, black sclera, orange pupils" tags. Same thing with the rodent queen and her "red irises, black sclera, red pupils" tags. AND IN FACT... as I show in a sample image, it is somewhat workable to do "blue irises, black sclera, blue pupils"!

Perhaps related to that is that sometimes, especially when generating close-ups or during upscaling, regular characters' pupils will be "empty" instead of being entirely filled in with black. This seems to be more of an issue with colors that got tagged with "eyecolor sclera" somewhere else -- so, for example, "orange eyes" plausibly got affected by Squealer's "orange irises, black sclera, orange pupils" tags. However, I didn't encounter this "empty pupils" issue all the time, and furthermore, it's trivial to fix, because the regular humans' pupils are so simple. So if that's the cost of getting Squealer's eyes correct, then it's worth it.

A complete list of tags used, ordered by decreasing frequency ("most tagged to least tagged"), is available as a .txt that you could edit to use for wildcards. Look for "Training Data" (or, thanks to CivitAI, "Training Images") in the righthand sidebar.

The .zip also includes the training settings I used (requiring 24GB of VRAM -- in my case, a desktop RTX 4090 card). I'm not necessarily endorsing my settings, but my Pony LoCons all use the same settings, so if you try ANY of my LoCons, you can see what those settings accomplished :)

_______

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 don't inpaint! 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, 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 OR TWO with "regular" upscaling, and I almost always end up at 2000-2500 pixels in height. When resizing for "regular" upscaling, 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).

___

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..

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.