>>455
i keep both eyes peeled but i havent seen anything dedicated to guro/snuff/ryona. you'd probably have to train something yourself. I've experimented with making TI embeddings but dont have anything worth sharing yet.

>>455
We need to organize it.
I have a massive pic collection but I need to sit down and tag them properly.


Any guide on ho to tag pictures for SD training?

found a model trained on IRL porn.

again it cannot do guro but can still be suggestive for captions

imgur.com/a/MMwZow2

>>465
Looks promising. Where to get checkpoint file, it it ever distributed?

>>464
You can use "interrogate" in Automatic1111 version or something alike to get initial tag set, then edit it to better represent image content.

Also, is there anywhere good place to post picture collection of 20-60 R18/R18G pictures if PIXIV is not an option?

>>468
Use mega or other filehost
mega may delete your stuff but probably it will not if nobody will notice and it is quite convenient

>>466
https://civitai.com/models/2661/uber-realistic-porn-merge-urpm

you can find the tags but putting images in this link to the webui PNG info tab

>>468
This is actually a good question.

Where is the new haven after pixiv was nuked?

found a choking LORA browsing pixiv: https://www.pixiv.net/en/artworks/105383393
huggingface link:
https://huggingface.co/yamack/lora-models/tree/main/choke

AyyMD fag here so while I can generate, unfortunately I can't train.

These aren't the best but I think more tags and/or more creativity from the AI might yield better results. No inpainting yet. Model suggestions welcome too, currently just using a hentai model.

(nsfw),(extremely detailed CG unity 8k wallpaper),(best quality),(ultra-detailed),(best illustration),(best shadow),highres,
(1girl),
(decapitation:1.7), blood, blood from mouth, corpse, injury, empty eyes, nipples, necrophilia, (wound:1.5), cuts, sex, human, underwear, weapon, bruise, (head wound:1.7), bullet hole
Negative prompt: (worst quality, low quality:1.4), robot
Steps: 34, Sampler: Euler a, CFG scale: 15, Seed: 2549232250, Size: 512x512, Model hash: 168144a879, Model: anyhentai_18

masterpiece, best quality, detailed, (1girl), (guro), ((ero guro)), gore, dismemberment, blood, blood from mouth, corpse, nude, empty eyes, nipples, vagina, necrophilia, (wound:1.5), cuts, sex, (death:2)
Negative prompt: (worst quality, low quality:1.4)
Steps: 30, Sampler: Euler a, CFG scale: 7, Seed: 68737965, Size: 512x512, Model hash: 168144a879, Model: anyhentai_18

(nsfw),(extremely detailed CG unity 8k wallpaper),(best quality),(ultra-detailed),(best illustration),(best shadow),highres,
(1girl),
hanged, suicide, decapitation, blood, blood from mouth, corpse, injury, empty eyes, nipples, necrophilia, (wound:1.5), cuts, sex, human, underwear, weapon, bruise
Negative prompt: (worst quality, low quality:1.4)
Steps: 34, Sampler: Euler a, CFG scale: 10, Seed: 732712575, Size: 512x512, Model hash: 168144a879, Model: anyhentai_18

>>479
Damn, these are nice. Can you share the model/prompts?

>>480
I made a Lora model trained on about 800 beheading images. It's very hit or miss right now but it has the right idea. I'm not sure of good way to share it

1. I can't believe gurochan is back.
2. I've been playing around trying to get some decent gore in SD.

>>481
catbox / anonfiles
or civit, it seems it not strictly against hardcore stuff, as long as it flagged as nsfw

>>484
Doubt Civit will allow something like that staying for long.

>>481
Nice. I can't fiddle with SD right now but it's nice to see a model exists somewhere.

>>485
Well... that was quick.
Lora got nuked in minutes.

https://anonfiles.com/qc89cea7z8/hunged_girl_safetensors

Sadly can not read comments... but i think is's something like GTFO-style.

>>486
>Endless hardcore porn.
>Tentacles
>gore gets nuked in minutes.
Damn.>>486

>>455
So no luck with a model trained specifically for gore yet?

>>490
Thank you, I will try it

Has anyone worked/found any shooting to the head execution models yet? Would like to have plesh :3

made this after some of inpainting and photoshop

Does anyone know what tags to use to generate severed body parts and open wounds?!

>>494
depending on what sd-model you use prompting "missing limbs","missing arms" "missing legs" etc may sometimes work for amputee stuff. URPM for example appears to kinda know what an "amputee" is.
Someone recently uploaded an Amputee - Lora to civitai; https://civitai.com/models/30632 ...Apparently the guy didn't upload any preview pics, so unless you have the link you won't find it on their search engine. Makes me wonder how many other relevant models lie hidden somewhere there.

>>495
>404
Anyone got it before it got nuked?

>>496
lol, that thing was up for a week + 4days early access. Only got nuked a few hours after I linked it here.

https://files.catbox.moe/liwwk4.safetensors

Don't know why catbox changes the filename. Original name is: "R18GAmputeeLimbless_v10.safetensors"

Only listed prompt was "limbless" iirc. You'll have to check your neg prompt to not contain stuff like "missing arms", "missing legs". Adding these two to the Prompt did seem to help. But don't expect too much. It makes either healed stump wounds or clean bandages and as far as I've seen only amputates the lower arms and legs, not the whole limb.

>>497
Thanks a lot, gonna experiment with it tomorrow.

>>498
post results.

For me it worked quite nice

https://www.pixiv.net/en/users/4750614
Bunch of R18G lora-s.

Though hosted on temporary storages like pixeldrain and anonfiles, so better grab it while it alive.

This is a LORA specially used for making impaled: nbbq. The attached pictures are all one-time pictures without post-production (except for mosaicing), and I think they are quite stable.

You can activate this LORA through [impaled by metal rod, impaled neck, impaled pussy], etc. There is a high probability that the stabbing on the head will fail, but as long as the posture is squatting and the legs are spread, the success rate of piercing the genitals or butthole is still very high.

https://drive.google.com/file/d/10wIXeH9GpTKpzWquliozroQNNj3x-VIy/view?usp=share_link
Lora: nbbq

After some more complicated img2img processing, it can also achieve good results.

aamzing.

what model do you use? ANythingV4.5 ?

Or something else?


I got pretty meh results ith Uber Porn since it has trouble removing the head.

>>511

I usually use kotosmix and AOM3 as checkpoints, and use beheaded LORA released by
>>36630, and beheadiffusionLORA from
https://www.pixiv.net/artworks/104649664 on PIXIV to make pictures.

>>513
this is photoshop with some img2img. I do not believe a txt2img produced it

>>516
Of cause it is INPAINTING, who said anything about txt2img?

https://civitai.com/models/44663/feetbox
this LORA is fun

>>518
Strange, civit agressively purge R18G, but this somehow escaped radars.

first try with sd

>>512 how are you getting the head on a spike afterwards though? inpainting or is it all in one prompt?

forest,girl, looking up ,nude, face up, (the stake pierces between the legs:1.4), ( stake out of the mouth:1.2), impaled, (open mouth:1.0), <lyco:impalement:1.6>, boobs, intricate details, cinematic lighting, stunning environment, ornate, finely detail,extremely detailed face,extremely detailed skin,(perfect body),(perfect proportions), (medium breasts), blush, blood on clothes, blood on face,
Negative prompt: bad anatomy, two heads, multiple heads, bad res, bad quality, error,malformed, art by bad-artist, bad-image-v2-39000, bad-hands-5, art by negprompt5, lowres, bad hands, text, error, missing fingers, extra digit, cropped, worst quality, low quality, normal quality, jpeg artifacts,signature, watermark, username, blurry, artist name, (worst quality, low quality, extra digits, loli, loli face:1.3), man body, mans, man hand, big hand, musculature, chair, stool
Steps: 50, Sampler: DPM2 a Karras, CFG scale: 8, Seed: 3459172844, Size: 512x512, Model: MeinaHentai, Clip skip: 2

>>522
Can you pleas link the impalement one? it looks fun to paly with

Also slaughterhouse looks like a nice location

>>523
https://pixeldrain.com/u/6TPd1hqt - for LyCORIS
https://www.pixiv.net/en/artworks/106978413 - author

1girl, (high selfie:1.3), young, ponytail, standing, pretty face, looking up ,nude, face up, (detailed mouth:1.2), ((the stake:0.8), (pierces between the legs:0.9)), ( stake out of the mouth:1.5), impaled, (saliva),open eyes, (execution:1.0), <lyco:impalement:1.6>,extremely detailed skin, (medium breasts), (blush:1.2), tears, (blood on face:0.4), (blood on stake:0.8),
Negative prompt: (worst quality, low quality, bad hands:1.4), bad anatomy, two heads, multiple heads, monochrome, extra limbs, loli, (text:1.2), (patreon:1.2), (logo:1.2), (signature:1.2), (watermark:1.2), (monochrome), (greyscale), door, table, stool, (stake on face:1.5), shelves, board,
Steps: 50, Sampler: DPM++ 2M, CFG scale: 7, Seed: 2285833074, Size: 512x768, Model: foxyansfw_V2, Clip skip: 2

https://drive.google.com/file/d/1yWiZsqBxlMVLzc_huq9-k0t9zi_NUFeQ/view?usp=share_link
Two-legged sheep LORA

Before that, I have been struggling with the inability to train a more convenient barbecue LORA, but yesterday when I was playing with automatic undressing LORA and Bijiri LORA, I suddenly realized one thing: If AI can be trained to recognize the undressing of the same person before undressing Finally, the concept of "the same character just changes in state" should be able to be trained before beheading and after beheading, before roasting and after roasting?

After realizing this, I tried my best to choose a headless barbecue with a similar body size, milk volume and height to be side by side with the standing drawing of the characters. I picked some materials with a small amount and filled in the tags to start the preliminary proof of concept. The result was unexpectedly okay.

This "before after" LORA is distinguished into three keyword concept sets:
limblessbody, body without limbs, the only keyword among the three concepts that can generate a body with normal skin color on the right.
hangmeat, hung meat, tied with ropes and hung in mid-air. It can be combined with bundled keywords such as bdsm, rope, shibari, etc.
impaledmeat, impaled meat, meat that has been skewered with iron or wooden sticks. Can be used with keywords such as iron rod, wooden stick.
Common startup keywords: beforeafter, pov, multiple views,
Keywords to generate meat with the butt facing the screen: from behind, ass, anus,
Using these concepts to transform and combine, this LORA will be able to generate a middle-section composition of "the character you input is on the left and the right, and a two-legged sheep torso will be generated that compares the character's size and milk volume on the left".

Increasing the weight of LORA will effectively cut off the head of the two-legged sheep. The recommended weight is <lora:beforeafter-v1:0.75~8>.
Weights below 0.6 will preserve the character's head, but usually produce a slightly weird hair color.
Since there are still many unknown effects, that's all I know at the moment.

Currently known problems: Insufficient training samples lead to too high repetition rate of some composition styles.
It seems that it is difficult to produce two-legged sheep with extreme body shapes such as big breasts and poor breasts.
The color of uncooked raw meat does not have a keyword, which leads to instability (maybe a new concept entry will be made in the next version? Distinguish between cooked and uncooked concepts?).
I didn't specifically set whether the right picture of the two-legged sheep after has a head. I guess I have to set a new concept for training.
In the future, it will probably be necessary to increase the upper and lower limits of the milk volume of some training materials, and refine the design and combination of some concept words.
Try it out and give feedback.

What's interesting is that when you pull the graph horizontally, more before and after partitions will appear. Usually it is dressed > undressed > beheaded and roasted. Isn't this very convenient? Although the frames in the middle are not very stable, usually only the leftmost and rightmost pictures can be displayed, but as long as the screen is wide enough, the whole process of dressing > half taking off > fully taking off > slaughtering can be realized.

Although I didn't train this effect specifically, I might consider strengthening this feature in this direction in the future.

>>521
just by accident lol
However, you can use the prompt to decapitate LORA in combination with impaled LORA to increase the possibility of this possibility.

>>527
It's amazing.
Thank you.

Nitpics. It failes making huge tits, Also realistc models tend to fail on it since I believe it was trained on an anime/cartoon dataset.

>>532
what tags do you use?

>>535

try 1:1 scale or wider picture

masterpiece, best quality, absurdres, wallpaper, 1girl, beautiful detailed eyes, light blush,steam,(gyaru:0.8),(blonde hair), blue eyes, single hair bun,one side up, blue scrunchie , (asymmetrical hair), kogal ,stud earrings ,(yellow sweater),(white school uniform), (shirt tucked in), (red bowtie), (blue checkered skirt), yellow cardigan,scrun,smile,<lora:galkoOshieteGalkoChan_galkoV20:0.9>,beforeafter, pov, multiple views, limblessbody, hangmeat,impaledmeat ,<lora:beforeafter-v1:0.75>, [ large breasts, pussy juice, shiny skin, curvy sexy body, ]
Negative prompt: (worst quality, low quality:1.4), (realistic, lip, nose, tooth, rouge, lipstick, eyeshadow:1.0), (dusty sunbeams:1.0), (abs, muscular, rib:1.0), (depth of field, bokeh, blurry:1.4),(motion lines, motion blur:1.4), (greyscale, monochrome:1.0), text, title, logo, signature, <lora:EasyNegative:1>, speech bubble, text, realistic,
Steps: 28, Sampler: DPM++ SDE Karras, CFG scale: 8, Seed: 3979926953, Face restoration: GFPGAN, Size: 768x768, Model hash: 49ef66fc4c, Clip skip: 2, ENSD: 31337

New version Before-after LORA V2

https://drive.google.com/file/d/1jJ1U9Xzj_r6JVbgLndrwssz44QJv4SYd/view?usp=share_link
The V2 version is trained based on the SD basic model.

https://drive.google.com/file/d/1A1qCtzGJfD6H3AnQsOEi7zKUGZnS-t7X/view?usp=share_link
The V3 version is trained based on the kotomix model. In theory, the human figure will look better. Use exactly the same training material as V2.

The revision of this version: Added more body shape changes, so that LORA can generate diverse body types such as big breasts, big buttocks, and flat chest loli.

Basic PROMPTS: beforeafter, pov, multiple views,
limblessbody: generate a body with normal skin color
hangmeat: generate cooked hanging meat
impaledmeat: Generate skewered meat

New concepts:
keephead: generate a head
decap: cut off the head
upsidedown: generate an upside-down female body

The three newly added concepts are a little unstable in practice, especially upsidedown is almost useless, it is difficult to produce pictures normally, and some adjustments are probably needed.

>>540
I love your model so much.
Kotomix gives me bad eyes, so I prefer using it with the realisian model.

french, long black ponytail, <lora:beforeafter-v3:0.85>, (hangmeat:1.1), beforeafter, multiple views, (huge breasts:1.0), submissive, (whore:1.3), slut, dominatrix, (fetish uniform), , 3 girls, shibari, sweating, same face, clone, roasted, vapor, bondage, white Shirt, limblessbody,
Negative prompt: fat, ugly, CGI, cartoon, comic, (poorly drawn hands:1.0), poorly drawn face, bad art, low res, bad anatomy, missing fingers, ((blurry)), (low quality, worst quality:1.3), child, 3DCG, drawing, monochrome, mask,
Steps: 28, Sampler: Euler a, CFG scale: 7, Seed: 2421811467, Size: 1024x664, Model hash: b7ebe4b20c, Model: realisian_v20, Denoising strength: 0.83

>>542
dammit I'm in love with this one when it generates 3 girls

>>542

Glad to see you all finding ways to tweak this LORA to suit your tastes. This also gave me the confidence to train other types of AI in the future.

>>546

Really appreciate it!

Found an Asanagi style on civitai
https://civitai.com/models/5941/asanagi-style-lora

works bad with most LORAs here except the hanging one.

but ith that one is quite nice.

Works nice with prompts as
business suit, school uniform, maid and glasses,

A textual inversion model works better hen combined with other stuff, but the girls are less attractive.

OP here. Follow me on Pixiv for R18-G loras and other stuff.
https://www.pixiv.net/users/4750614

I know some inflations lora's but they seem unable to generate stretch marks and bursting guts.

New LORA release: separation of body and head
https://drive.google.com/file/d/1pS8jcEMp8bmMEM3cgNymbmkJjCRgzQgF/view?usp=sharing
download link

The basic effect is to generate a girl's appearance before she was alive on the left side of the picture, and generate a composition style of her kneeling or lying on the ground after being beheaded on the right side, with her head rolling down beside her.

Basic PROMPTS:
ghead,
severed head,

buff PROMPTS:
blood,guro, death, ryona,trembling, kneeling,seiza,spread legs, lying, on back, defeat

Even in the case of 1girl, solo, as long as the canvas is long enough horizontally (about 4:3 or more), there will be a time-series change diagram of being beheaded gradually like the two-legged sheep Lora.

Currently known problems: The probability of generating heads is unstable (30% probability is only headless corpses, and the headless rate is higher when there is a character LORA, the reason is unknown). It is recommended to increase the weight of severed head or CFG. I have not specially trained hands and feet, so once Lora intends to draw hands and feet, there is a high probability that he will not draw them correctly. Either the left and right feet are in the wrong direction, or they can't tell which side is inside and outside.

Welcome to report your experience or suggestions for improvement.

>>560
good work

>>455
>>560
This is amazing. Thanks for the model¡!!!!!

trying to mix some other LORA, some failed some okay.maybe needs more training picture.

All this is good, but I'm awaiting a more general bloody wounds/ryona/injury/bruises kind of Lora. I tried to train one myself and it was awful.

Im a complete novice to anything AI, and wouldnt even know where to start.

Surfing through most threads, seems depending on which one is used you can generate different results.
Like theres the more anime/cartoonish ones, and then the more realistic ones?

I'd most likely want to experiment with skull-pens , ears , eyes, forehead, back of head, very top of head. These types seem to be lacking.

Its getting it right spot on with the impales and sorta feel my guro creativity spark igniting again lol.

Soooo Any advice?

Thx everyone and keep it up!

>>565
If you are really starting at zero I'd suggest you watch a few basic tutorials for the use of automatic1111's Stable Diffusion local install. Get familiar with the txt2txt basics, inpainting and the use of loras /lycoris for a start.
You'll find tons of base models and loras/lycos ranging from photorealism to anime on civitai.com. No r-18g there other than some tame vorestuff and a wee bit of blood but plenty of usefull stuff nonetheless.

Most r-18g loras can either be found here or on pixiv. (just search for "r-18g lora") There are some hidden gems though that aren't tagged properly.

As far as I know there isn't any straight forward lora for skull- or eye- penetration etc. However there's a lora that might be of interest to you here: (scroll down to the "head destroy lora" entry) https://monster-maker.fanbox.cc/

I haven't gotten it to work for txt2img, but it works very well for inpainting. it basicly allows you to put holes into the head, and generate some nice brainsplatter. You should be abled to produce some nice entrance holes or even brain penetration...

Another hidden lora-gem can be found here: https://kemono.party/fanbox/user/24196387/post/5655457 (I just grabbed the link and then Kemono went down. If it still doesn't work in a day or so I'll upload it here) It's a dissection lora. Not the prettiest but so far the only one I know of.>>565

>>566

ooops, I meant txt2img basics, not ""txt2txt"... >.<

>>564
this one's pretty good at blood
https://civitai.com/models/40921/zombie

>>473
Anyone figure out what the keywords are for this? The hands look awful, but otherwise its good.

does anybody know what github or sites does this? please and thanks

if someone please link them thx

>>538
do you have a link for v4?

>>573
https://drive.google.com/file/d/1ESRYPXvLeVGSTSZOxv_P705swU5WYt_1/view?usp=sharing

Use it if you need it. It just deletes all the keep head concepts that have heads and may interfere with the generation effect, and adds some training materials that strengthen the sense of flesh. In the future, I may switch to anythingV3 and try a new version.

>>574
thanks. i really like the meat tags

https://mega.nz/file/U1FSSR4Z#oZePQ9pgwRreZ3bAX_mU1rFMjFagXVXniiqyv0As4vg

"corpse display"LORA made by pixiv users/20344975 Deawfwaef, it's pretty good for me.


masterpiece, best quality, absurdres, high resolution, extremely detailed,
【品質控制】

1girl, solo, <lora:lying:0.65>,dead_lying, expressionless, corpse ,cum in pussy, pussy juice, nipples, crying with eyes open, blood,
【Deawfwaef的屍體展示LORA，相關表情與動作的控制詞】

<lora:redcoat:0.6>, narrow waist, large breasts, blonde hair,long hair, blue eyes, boots, ww1era,khaki uniform,equipment layout, bag, pouch, peaked cap, torn clothes, military uniform,
【英國陸軍制服LORA與配件相關控制】

Negative prompt: (worst quality, low quality), bad_pictures, negative_hand-neg, text,easynegative, badhandv4,
【通用負面控制】

Steps: 48, Sampler: DPM++ 2M SDE Karras, CFG scale: 7, Seed: 599382441, Size: 1344x896, Model hash: ef49fbb25f, Denoising strength: 0.5, Clip skip: 2, Lora hashes: "redcoat: c63681b958a4, lying: 50629f7e5126", TI hashes: "easynegative: c74b4e810b03, badhandv4: 5e40d722fc3d", Version: v1.5.1

need large image to make face and equipment more detail.

"cannibalism"LORA made by pixiv users/20344975 Deawfwaef. it's awesome.

masterpiece, best quality, absurdres, high resolution, extremely detailed,
【品質控制通用】

1girl, solo, food,cry,apple,fruit,object_insertion,chicken_\(food\),plate, dinner, drinking glass, wine,<lora:re-eat:0.75>,meat, embarrassed, crying with eyes open, shy, full-face blush,
【秀色LORA相關控制】

hyouka, <lora:hyouka_offset:0.5> , black hair, long hair, hairband, open mouth, anime coloring, purple eyes, hair between eyes , medium breasts, <lora:fuyumi:0.7>
【人物LORA相關】

Negative prompt: (worst quality, low quality), bad_pictures, negative_hand-neg, text,easynegative, badhandv4,
【負面通用】

https://mega.nz/file/8gd0DbpY#kIuRPiFwu-L_00xOBXIMglySQeHd4-2cRxJamS6RLpE

heres "cannibalism"LORA

>>560
>>558

took quite some time to make it work. It doesn;t synergies with other Loras well.

But the results are amazing when it works.

>>580

yeah, maybe only 50% success rate, but still works.

Is there a link where every single lora posted here is compiled a one place? Right now it's everywhere.

>>583
as far as I know no.

I can make a link with all of my LoRa's but it is 20 GB, and most of them are useless.

does anyone have this lora its on bido so cant get it https://www.pixiv.net/en/artworks/110476989

>>585
https://files.catbox.moe/fb0syw.safetensors

Catbox randomly renames stuff I upload. Original filename is "ci-fu.safetensors". Enjoy!

>>584
I mean a rentry or something similar that list off the download links to your LoRAs

Is there a lora where the dead body is placed on a table in a morgue?

>>587
we need to make a cheap knock off of civitai.

rentry is good but it doesn't allow comments and images.

We will need some way for people to ocmment and add their Lora when I will be too lazy to mantain it

I wonder if there are some good loras for wounds intestines and fresh amputations in more realistic style?
Something what could go well with 3D renderings

Does anyone know of/can I request a defloration/virgin blood lora?

Should be simple, but I haven't seen anything close since Loras went mainstream

Is there a lora that can color the blood whatever you want?
I want black blood for purposes.

>>592
Don't know of a lora for that yet. You can always make one or just drop the image into photoshop if it's only for 1 image. Sorry I can't be of much more help

A new SDXL model came out
https://civitai.com/models/257749?modelVersionId=290640

Do not be fooled by ponys, it can geenrate AMAZING hentai images.

>>594
Any tips on using SDXL? I've never used it before and I heard there's alot of differences from 1.5

>>595
In general SDXL preferes natural language over the key word spamming of SD1.5 It is not as agood as Dall-e 3, it is in between.

Now the model linked (pony XL) have been traine don danborour. So it doesn;t like natural language, it likes the spamming of tags from danborour + some custom tags int he description of the model.

in general FInd an image on cvitai you like made with that model, read the prompt on cvitai, copy paste it in your own system and then play with it.

Good luck.

For picrelated the pormpt was

score_9, score_8_up, score_7_up, score_6_up, police uniform , nipple piercings, animal collar, barcode tattoo on her pussy, submissive, , slut, fucked , sex, whip marks, asphyxia, choking, rape violent, brutal, missionary


and negative

3D, CGI, render, photo, text, watermark, furry


With Pony Diffusion XL and smooth anime style lora

>>597
what do you mean? He wrote his prompt here >>536

test

I tested the SDXL hanging lora from the rentry and it is very good.

Unfortuantely it doesnt work with anime, realsitic images, and is bad at faces.

but hey, it is progress!

lora file: https://files.catbox.moe/j1xppc.7z
a dissection and dismemberment lora for SDXL, works for realistic model only.
example imgs: https://files.catbox.moe/8d96cs.rar
prompt useful: dead body, blood, anatomical dissection, lying

test

>>601

It's awesome. Added it to the rentry. Are you the author or there is aocmmunity out there that I missed?

https://rentry.org/R-18G_Stable_Diffusion

>>603
ye i am the author.
here is another amputee lora i made for realistic SDXL model: https://files.catbox.moe/ct6ty6.7z

example imgs with metadata: https://files.catbox.moe/5wx8k2.rar

prompt useful: amputee, stand on ground, sit, lying

weight: 0.5+

note that if you see some color block spot in the generation, you might need to change the sampling method or lower the cfg scale.

>>604
Wow this works amazingly. Even interact well with other charcter loras.

Where do you usually post? besides here. If you don;t want to post gurochan you can contact em on my pixiv

https://www.pixiv.net/en/users/101800918


Unrelated attached pic!

Rentry updated .

Soem cannibalism LORA added.


Also a fun lora attached.

We are really missing a decapitated LORA for Pony.

BTW you can now train Pony LORAs directly on civitai.

In case you pC is bad you can do it for free.

a decap lora fro pony appeared on civitai.

just search Disembodied head

>>515
how do you do neck slashed and hand holding head and pulling head like that on dreamshaper,dynavision, sdxl or similar? i have had little luck

>>608
I'm not that anon. The pic you linked was done in SD1.5

so I have no idea how to do it in SDXL. Pony is quite NOT flexible, so unless you fins a throat slashing loras is gonna be bad.

on SDXL you might have better luck when descrbing it explcitly.

What are you current results in that front?

>>609
fins?
whats that


well i describe it very very explicitly sometimes repeating the same sentence several times

but the result tend to be very mildly correct or straightforward confusing.

i am outta clue already.

>>610
>fins?
I wanted to write "find"

So basically my answer is: I have no idea.

BUT someone on pixiv made a beaheding lora for Pony

It was added to rentry
https://rentry.org/R-18G_Stable_Diffusion

Among a bunch of other pony loras.

I made a LoRA based on AyaSwan works.

Yo ucan find it at the bottom of the rentry here >>611

It works very well when combined with other loras, here it is combined with headtoy lora, both available in the repository

>>613
I did an autoamitc tagging through DeepDanbooru, so it is not very good at understanding prompts.

Might revisit it int he future with better tagging.

But it is 2000 images. It is too much.

it can generate nearly good images, but it does need proper tagging.

You must verify that the girl is dead.

Do you take requests?

>>629
Yes. But please submit to my thread, so we do not derail the current thread

My thread is here >>39852

>>611
Could you please share the promts you used for the "thumb in mouth" move? Btw, this lora is great, but somehow I can't get clear neck wound. Are you using any artstyle lora?

I really can't make these loras work. Can someone provide box or an upload w/metadata so I may examine?

Is there any (non-realism) Lora for pony for at gunpoint and/or gun related violence? Seems to be extremely difficult to get a character to even hold a gun let alone aim at at another character.

>>518
What prompt did you use to generate that?

>>634

example:

1girl,solo,the wall photo have a girl portrait,in photo,pov,feetbox,barefoot,only head,(severed_head:1.4),svd girl,box,soles,girl boots,no body,no arm,no hand,black socks,steam,hot,wet,<lora:feetbox_v10:1.1:lbw=MIDD:step=1-28>,<lora:0iggb5:0.8:lbw=MIDD:step=1-28>,

<lora:shi2:1:lbw=OUTALL:step=1-28>,blonde hair,blunt bangs,long hair,blue eyes,pointy ears,peaked cap,f((id card,military equipment layout,sweat,steam,map,paper,olded military uniform,medals on floor,photo \(object\),)),

Where can I find the contents of “AyaSwan Style v2”?
I would like to check the prompts to make tentacles or headless necrophilia, but I don't know how to do it.

I cant find: Wall_of_meat.safetensors people are using. can someone please throw it up on mega.nz then add it to the rentry?

>>611
Holy fuck that's hot! That's a new favorite right there!