I'm sticking with obvious VG/Anime characters so I hopefully don't step on any realism rules, and that's my preference anyways. But you can get liveleak quality horror if you try.

If you want realistic characters though, your best friend is "looks like [character]", and the prompting formula I've found gets great results is easy and simple.


>Photo of a woman [whatever is happening here].
>The woman looks like [character], with [a quick identifying feature, like hair] and wearing her [whatever clothing items].
>The background [flesh out the scene].


-----
Raw photo of an athletic young woman falling back against a stone wall inside an ancient mayan temple. Many small spears jut horizontally from the wall, impaling the woman. The spikes emerge from her torso, dripping with blood. The woman's mouth is open and her shocked eyes wide in a horrified expression. One of the spikes emerges from her mouth, blood running down her chin. Her hands are open and tensed tensed.

The woman looks like Lara Croft from Tomb Raider, with a tight blue tanktop over her ample breasts, pistol holster straps, and her hair in a ponytail.

The background is the carved stone wall of a jungle temple, with crawling vines and moss. A large circular mayan pattern is carved in the stone, partially obscured by the woman. There are a few blurred spikes in the foreground and some behind the woman, adding depth to the image.
-----

The long prompt is just for detail. It knows most of these subjects from a single word. It knows "impaled" without trouble. Only loras are the filter bypass and "Realism Engine".

-----
photo of a slim woman impaled vertically through her chest on a large sharpened wooden rod emerging from a bayou swamp. The girl looks like Jinx from league of legends arcane.
-----

For you trophy people, it knows what "taxidermied woman" is without any help.

-----
Photo of a woman taxidermied. She is naked and posed standing holding her Sheikah Slate to her chest in the corner of an artificial display of Hyrule Castle stone walls, in a display case. The display is behind glass, and a placard on the glass reads "Zelda Taxidermy". The back of the case is a matte painted background of Hyrule.

There is a seam of stitches up the middle of her torso. Her eyes are fake, made of glass.

The girl looks like Zelda from Breath of the Wild, with blonde hair, wearing her boots and gloves, with elf ears.
-----

Very freaky it can do this. I cannot imagine how or why it would know how a taxidermy pussy looks like. Sorry if this one is too freaky! Again only filter bypass and Realism Engine.

-----
Photo of a woman's head taxidermied. The head is mounted to a large plaque on the wall in a hunting lodge. Below the head is a vulva taxidermied. The taxidermied vulva is mounted to the plaque below the head.

The head looks like Princess Peach, with blonde hair and a crown. Her skin has a pale, dry texture of taxidermied skin. Her eyes are fake, made of glass, glinting in the colored lights.
-----

Anyone like the retro film? This one has a single blood lora I found.

-----
raw movie film still, live action, photo of a fit, toned woman impaled vertically on a long wooden pole in a grasslands field at night. Her feet dangle limp above the ground.

The girl looks like Casca from Berserk, with a tan complexion and short black hair, wearing her torn thigh high boots and a piece of armor on her shoulder. Blood is splattered on her body. The top of the pole is sharpened into a point.

The sky in the background is an ominous solar eclipse, with a total eclipse casting the scene in an eerie red light. The ground is scattered with the remains of a large battle, with dead knights and horses scattered across the bloody field, with and broken spears and swords jutting from the ground. Retro film still, movie scene aesthetic.

>>22142
Blood lora made from Terrifier movie images apparently:

https://huggingface.co/bobo1666/ge/tree/main

If you have tried to make someone upside down before, you know how difficult it is. Almost as impressive as knowing Kemono Friends. Blood lora again at .4 strength.

-----
Amateur vacation photo of a hunter in the african savanna under a large tree. He is standing next to a naked woman that is hanging upside down from a rope around her ankles, she is suspended above the ground from the rope around her bound ankles, her arms and hair dangling down. The man is gutting her body with a small knife, pulling the organs out of a vertical incision in her torso. A pile of viscera an intestines lays on the ground below her. The picture has film grain and looks like an amateur shot from the early 2000s.

The girl looks like Serval from Kemono friends, with natural blonde hair and serval ears, and animal print stockings and long gloves.
-----

On the other hand this one was not easy.

-----
eye-level photo of a woman. She is in behind a table with her large bare breasts laying on a plate on the table. The table is wooden plank with nicks and scratches. Her hands are down behind her back.

The woman looks like Nami from One Piece, with natural short orange hair and a tattoo on one shoulder. She looks down at her breasts with a surprised and slightly anxious look, biting her lip.

On the left of the image, out of frame, is a man. He reaches out and grabs the girl's closest breast with his hand. With his other hand he cuts down into the girl's breast with a carving knife. The knife is horizontal, slicing down into the breast, the knife is half-way in the breast slicing it deeply, revealing the fatty meat inside of her breast, minor blood running from the vertical incision. Salt and pepper sit on the table, and there is a carved Devil Fruit slices on a small dish.

The background is a pirate ship. A man who looks like Luffy, wearing a red shirt and straw hat, can be seen in the background, smiling and giving a thumbs up. The aesthetic is Pirates of the Caribbean film still style, live action movie.
-----

The Realism Engine lora fights the spider web, trying to turn it into bondage, but it mostly works.

-----
Raw photo of a naked woman bound in web, cocooned in spider web and suspended horizontally, her arms and legs tight together, inside a large cave. The background is a dimly lit cave with no light source except a torch laying to the side on the floor, casting flickering fire light into the pitch black cavern. In the distance, a giant shelob spider looms. The scene is a film still from Lord of the Rings, with dramatic cinematic lighting and film grain.

The woman looks like Aqua from Konosuba, with light blue hair, wearing her thigh-high boots mostly covered by a thick wrap of webbing. Her medium breasts bulge out from the webbing, exposed and hanging.

She is completely wrapped in web from head to foot, encased in a thick cocoon of web. Her expression is an exaggerated scream of panic as the spider approaches.
-----

It also knows so many movies and TV that you can easily ask for them.

-----
Scene from TV series The Walking Dead. A woman is on the ground, panicked and struggling. A zombie pins her down, biting her breast and ripping the flesh with its teeth.

The woman looks like Claire Redfield from Resident Evil, with brown hair in a ponytail, wearing her red jacket. She is struggling, propped up on one elbow on the ground, gritting her teeth in pain with wide, shocked and fearful eyes looking at the zombie. Her jacket is open and her grey shirt is ripped and torn open at the front, revealing her breasts.

The zombie is a rotting corpse with mottled torn flesh. It holds the girl's wrist down as it leans in, its teeth biting into her breast and ripping raw flesh and tearing skin, with and dark, viscous blood.

The background is a rainy alleyway at night in New York. The girl and zombie are struggling on the wet ground, dirty and splattered with blood in the dark night. Her hair is damp and her outfit wet with rain.
-----

It knows exactly what hanging is without help, and it is so good at hanging that it is the first image I'm really not sure I should post. It's so real it makes me sad for a Miku cosplayer that doesn't exist.

Here is the prompt though, short and simple.

-----
raw photo of a naked woman hanging from a rope around her neck. Her feet dangle.

The woman looks like Hatsune Miku, with natural teal hair in two long pigtails, wearing her thigh high boots.
-----

It will take tags as well, but it wants to force them to be anime. I had to tell it to look like CSI to make realism, and still her hair is almost anime LOL not bad though

-----
raw photo looking down on a deathly pale woman lying on a metal table, her eyes are cloudy, morgue, parted mouth, looking up, blood on body, wound on chest, wound on face, blood on face, gore, autopsy, her chest cut open, a long gash in her chest, she is being dissected and her torso is cut open, her organs and guts visible,

The woman looks like Yor Forger, with black hair and a golden hairband with a golden metal flower decoration on the side, wearing her gold earrings and an elegant black dress ripped open to reveal her body. The neck band of her dress is intact.

Film still from CSI TV show, with 2000s TV style and film grain. Dim blue clinical lighting.
-----

Never thought we'd get a Pony replacement, but here we are.

-----
Photo of a woman bound by vine tentacles, her arms and legs spread, in an alien jungle with sci-fi plants like a scene from Avatar on Pandora. Tentacle vines wrap around her wrists and ankles, pulling them apart as she struggles suspended in the air, one arm pulled closer against the tension and her legs straining, one leg partially bent and the other extended. A tentacle vine lifts her skirt, revealing her vulva. Another tentacle emerging from the jungle foliage rises up and penetrates her vagina. She has an ahegao face expression, head tilted back with eyes crossed. A tentacle emerges from her mouth, splattered with blood, and ends curled in the air. Blood runs down her chin onto her clothes. Blood drips from her vulva.

The woman looks like ahsoka tano, with natural orange skin, wearing her headpiece, boots, vembraces with fingerless gloves, and a her torn blue outfit and skirt.
-----

i started using Krea2 as well as one part of my creation process, very fast and looks good. My own observation, Krea2 is very prompt dependent, unlike my favorite "illustrious" and it's forks which can provide lot of content without directly specifying it. I see some people using local "prompt enhancers" but i find them slow because they also using AI models for this enhance, but instead i would recommend Grok - he is super smart for creating prompt for OTHER models, not just for Krea2.

>>22152

Definitely true and good advice. Unlike illustrious it will not give you much randomness or create details you do not ask. I don't mind that as I think "prompt crafting" is kind of fun, but many enjoy rolling the dice. I know using base Krea2 and a turbo lora is supposed to provide much more of that chance than Krea2 Turbo. I haven't tried that method yet.

>>22149
Thank you for making/sharing this. It might be one of the best images I've ever seen. I got instantly erect when looking at this. Highly attractive woman with beautiful pasty pale skin, dark hair and blood dripping from her mouth. The clothing she's wearing is particularly attractive as well.

>>22154
I forgot to ask if you made or have any more similar images. Is this one from a set?

amazing, all of it 👍👏

>>22137
Can you share your workflow

i have tested all your (fellow anonymous) Krea2 prompts and most of them are working, even i have "unlocked" krea2 some of this samples refuse to produce gore and blood, and some does and some of them does it very weird, but after applying your lora it started to work properly,

like OnePiece it was just poking unharmed breast, with your lora it start making actual cuts. but the taxidermy i was not able to reproduce - i still just look like a paper toy :D kinda hilarious.

Then another problem, the Lara image, all the spears are reversed, i had this problem with Qwen Rapid before, knifes, needles arrows spears are reversed, looks like same problem same here.. if you know a workaround tell me.

As for zombie image it reproduces exactly the same, you know it's very easy to paint body just with blood even before Krea2 exists, if you saw my other thread i made many images about zombies/ghouls, the most recent one fallout vault girl surrounded by ghouls, i had another image after that where ghouls tearing her apart and eating parts (not just touching body with teeth) and it wasn't good... models still can't do that a proper dismemberment, it had to be heavily shopped.

and thank u for sharing this, with krea2 next level is definitely unlocked, it surpassed the z-image.

>>22137

Can you share your settings?

i don't have original author workflow, but it's pretty simple.

>>22162
this is not because some special settings, it's because of the model checkpoint, if you use "krea2_turbo_fp8_scaled.safetensors" even with blood\gore lora it will not work. you need to download uncensored checkpoint first, and this model is hot right now, it was updated again just few days ago. these models are publicly available here:

https://civitai.com/models/958009/
^ the page also have "uncensored" z-image and many other top models.

https://civitai.red/models/2242173/
^ this one something new, haven't tried it yet.

also because this model is pretty new, old Comfy will not work, you need to download latest up to date comfy first, i recommend do it in separate folder, keep your old one too because new comfy may mess up old workflows and custom nodes.

>>22162
I am glad they are (mostly) working! I will add that I am using this filter bypass lora. I found it works much better for unlocking violent content. This and Realism Engine are all you should need. Even the blood lora is just for extra effect. I'll share a workflow later when I can.

https://civitai.red/models/2746817/krea2-filter-bypass-fedor

>>22155
I'm glad to see a Yor appreciator. It is just something that I made for example. But perhaps I will try more...

>>22169
ok with the "krea2-filter-bypass" the full gore is unlocked, as well as pussy taxidermy (how the hell it knows that?), it's just ridiculous, this is not edited image, 1 click gen, it can slice, chop, and knows all internal organs. that's it guys, free uncensored local max gore unlocked, there is nothing to photoshop anymore.... :\

And for those who say "omg photorealistic! this one is more photorealistic than other omg", guys, it can do anime too, it's just one word difference in prompt, just look at this.
(don't mind small glitches, it's unedited images 1 click stupid gen).

>>22170
>>22171

Now we're talking! Wow, I didn't even expect it could do that much. Holy crap, man. There's hardly even a need for loras anymore, but if people do start refining it with niche loras it'll be unbelievable.

>>22163
>>22159

I don't know if uploading an image here will strip the workflow. We'll find out. You should be able to drag this image into ComfyUI to get the workflow. If not, here's a workflow.json on the first free file host I saw. 6 day time limit though.

>>22200
link, because i am dumb and left it out

https://filebin.net/58mjtqt2gc8tidg3

most simple workflow for 30xx

it's too bad Krea doesn't have an image 2 image. there are a couple of workflows that claim to do it but they are best effort attempts, on the bobo1666 ge model though kudos for making that, I can tell its from the Terrifier for sure. and excuse my ignorance, but what's the difference between the 1750 and 2k model, seems the 2k version just Cronenbergs renders into The Thing. is there a high low to shoot for or a combination of the two. Also drag and drop of the tentacle photo did carry over the workflow, just fyi, cleanest Krea2 workflow I've seen to date. I'm still fighting with trying to get up scaling to give me an output that looks as crisp as say the lara croft image, but that's on this end I'm sure.

>>22238
Devs say they're working on Krea2 Edit edition right now, which if so will be insane. Grain of salt though because Z-Image said the same thing and never happened. As for the lora I didn't actually make it, so I cannot say for sure, but I think 1750 and 2000 are training steps. So the 2k model may be "overcooked". I use the 2k lora but on maybe 20% for most things to keep from making flesh horror. Glad the workflow works too!

Loras can also affect image quality, so if the Terrifier lora was trained on less than HD movie stills it might be making quality lower if it's too high. This is using the workflow from above with Bypass at 1, Realism Engine at 1, and Gore at 0.2 and I thought it came out pretty crisp.

also they did something with the model, now they offer "Krea2RedMix-10Steps-fp8-scaled-ComfyUI.safetensors", but before it was another one and it was better quality, with some other glitches tho, current one renders different (still with gore yes). anyone know how to download previously uploaded models from civitai history?

Doesn't quite know how a neck cross-section would look, but you could probably fix that with better prompting. It knows a *lot* of celebrities. They really don't care about likeness and copyright lol...

>>22292
I think civarchive keeps and organizes everything with available download links if you know the model you're looking for.

https://civarchive.com/search?base_model=Krea+2&rating=all&type=Checkpoint

>>22295
thanks, i will try dl there previous versions.

also after some research i think there is more complicated system, krea2 is not just trained on gore images. like the old models used to slap meat everywhere, no. here is some sort of anatomical multilayer system, each layer can be on and off. here i tried turn on some of them and can see now it displays good knowledge about vessels and muscles. only if we can find names of all layers and access them

another proof of multilayer system, when everything is off still can clearly see shapes of a girl's body, krea2 is very advanced.

can anyome make a workflow how to do videos after creating an image?

>>22291
I'll hold my breath on the I2I or edit then. What was your prompt to get the rib cage to show, all I get are weird fleshy lines. I got favorable horror movie output using 1750 set to about .6 so may have been blasting the 2K lora

>>22302
You can build a workflow that then feeds the output image into something like wan or ltx, but the easier method would be to just generate and review your batch of images, select the one you want, then send it into wan with their simple I2V model. I think you are looking at way to much processing time to build it all into a single workflow.

>>22299
That looks crazy, is that using the realism Lora? and then did you just tell it give a "layered" biology text book image or did you have to tell it "from shoulder to lumbar show nervous system from lumbar to quad show musculature"?

>>22299
That's wildly impressive.

>>22307
Prompt below. As Mr. FleshReaper says, it's very detailed and specific, so if you want a detail you have to ask for it. "two bloody wounds" will just get you generic meat texture. "showing ribs" added ribs only and you could see into an unrealistic hollow body inside. Adding "and muscles" makes them look like ribs with more natural muscles. Also worth noting that I use the original Krea2 model, not one of the ones trained for porn. Those can sometimes overwrite the model's ability to do certain niche things in favor of expanded NSFW knowledge.

-----
Raw photo of a 22-year-old woman fast food worker holding a tray in a McDonald's restaurant. She is nude, wearing only a McDonald's employee hat. Both of her breasts are missing. Her chest has two huge circular bloody wounds, showing her bloody ribs and muscles with flesh removed on each side of her chest.

On the tray are two small severed breasts lying up on the tray in a small pool of blood, nipples upwards.

The woman looks like souryuu asuka langley, with natural red hair with bangs and twin tails under her hat. Her head is angled slightly down towards the left and she has an exaggerated angry pout expression, eyes looking away, frowning in annoyance.

In the background, blurred and out of focus, a fast food worker cosplaying like ayanami rei, with natural short light blue hair and a McDonald's hat, works the cash register.
-----

Can it do hanging? I gave it a shot and can't get it to get away from normal bondage.

>>22439
"photo of a naked 22yo woman hanging from a rope around her neck. Her feet dangle." + scene flavor is all it needed.

If you use a custom version of Krea, like "Dark Beast", it may be trained on so much NSFW that "hanging" has been overwritten by bondage using similar words. "spit roasting" for instance will give you something impaled on a spit in base Krea2, but will give you the sex position in some custom versions.

oh shitt.. damned true. krea2 is something. wtf thing they trained to get this shit.

love it.

never use the garbage tier shitmixes you find on civitai, especially the chink made ones. Always use OG + Loras or Full finetune ( which currently doesn't exists and probably never will )

>>22466
Awesome pic, and they Krea2 does a lot of things out of the box.
Where did you get cutifier_krea2.safetensors?

I find it very inconsistent.
I make a prompt, it produces image. I change tiny details on the prompt, it produces a black image.
I thought it might be related to age of the subject, but not even that seems to be the case.

>>22473
civitai just search cutifyier

>>22481
which checkpoint you use? i never got black image with base gguf quant version. and always crash oom using mix version when do more than 1 batch image. what i know is black image sometimes cause of sage attention. need disable it. but never heard of getting black image just by change prompt. what prompt u use then? to see if i get black image too or not

>>22483
It was sage attention...

It does furries really well!

SageAttention is fucking shit, stop using it. It is only useful if you do 50 steps on a 720p video model so it won’t take you 3days to complete. Imagine using a model already destroyed down to FP8 and then slapping onto it an attention mechanism that quantizes it even more..yuck.

Holy shit you guys were right.

More

More

Who knew 3kb of lora could ever do this lmao

same problem with pencils alignment, happens with arrows spears blades etc, any one know how to fix?

>>22679 I'm curious what prompts you use to animate that.

>>22690
It's Wan 2.2. just say "man thrusting penis into dead girl's mouth. She doesn't open her eyes" and it'll pretty much be enough

>>22678
Can we get this one animated

>>22689
Honestly really impressive how bro made his phone disappear.

>>22686
Just gotta think like AI. A spear is a long stick with a sharp point bit on it. You asked for a spear, not a long stick, so it shows you the point to prove it's a spear and ends up drawing it backwards. So don't ask for a spear, ask for someone impaled by a long staff. If the spear tip is hidden anyways it may as well just be a stick to the AI.

You could probably force pencils to be the correct direction by mentioning the eraser. Same deal.

>>22142

how much lora strength did you use here?

i used 10 eros to animate a necro scene where the girl's face and eyes doesn't move and it won't let me.

which animation model do you guys recommend for a 12gb VRAM and 16Ram? that is fast and follow prompts nicely?

people told me about Sulphur 2 and Wan 2.1 or 2.2.

any help would be appreciated

>>22739
1.0 for filter bypass and Realism Engine, then 0.4 for the Terrifier lora. Specifically the 2k version:
>>22143

>>22741
I've used Wan2.2 successfully with those same specs and default workflow on ComfyUI. I had to use gguf q4, but quality was fine. You need to use loras and creative prompting to get anything but vanilla sex though, and it takes like 8 minutes for 5 seconds of video... Still, if your specific taste is for necro look up the "Timefreezes" lora. It's probably exactly what you need.

https://civitai.red/models/2437696/timefreezes

krea 2 is kinda weird lol but yes insane gore

Awesome image! How did you get the rope to cinch up on the neck like that? With all my gens the rope is around the middle of the neck and I've been searching for a more Elirawne look on the hangings.

Believe it or not I just kept telling >Deepseek< to rewrite the promt to emphasize the tighten rope like 4 times.

>>22751
Instant photo, harsh direct flash, heavy film grain, blurry, out-of-focus edges, washed-out faded low-saturation colors, yellowed highlights, authentic vintage Polaroid snapshot frame. Erotic horror.

Realistic organic blood, not glossy or thick, pale watery crimson, faint smears, dried brownish edges, messy stains soaked unevenly into skin and fabric, no plastic shine.

A Chinese woman cut in half cleanly at the waist, lower half detached and lying separately about half a meter away, legs spread wide, pale skin, realistic wound at the severed waist showing subtle dark red muscle and fat layers. A small amount of intestines spills out from the lower half's open cavity, dark crimson smearing the flesh, not excessive.

Her upper half lies twisted on the dirty floor, still alive, hands weakly grasping at her spilling gut, trying to push it back inside. The severed waist shows a clean surgical cut with dark subtle bloodstains and minimal gore, raw but not cartoonish. Her face is frozen in pure visceral horror, eyes locked on her own lower half, mouth open in a silent scream, a thin trickle of watery blood running from the corner of her lips down her chin.

Dim grimy room, stark contrast, nightmarish surreal atmosphere. Only the two separated parts of the same woman present in the frame, no third figure.

Would you mind posting the prompt for the hanging pic if it isn't to much trouble? I keep trying to no luck.

>>22762
Unfortunately due to a malfunction it is no available to me anymore, you might just have to figure this one out for yourself.

No problem. I'll keep trying, but at least now I know it's possible. Thanks!

>>22748

thank you for showing me this lora.

could you PLEASE share your workflow with me?

>>22777
Thank you, kind anon. I am glad to share. ComfyUI Workflow in image. Drag it into ComfyUI and it should load automatically:
>>22200

Also for anyone who enjoys horror vibes in your guro, this lora is a great find. The Nightmare variant is more gory, the Fearbound version is more creepy.

https://civitai.red/models/2749232/obsidian-nightmare-krea2

>>22811
nice detail

>>22811

you're so kinda, also i was talking about the workflow where the 2.2 wan time freeze actually work with nsfw.

>>22777

also the provided image is in jpeg, i couldn't get it to work in my comfy.

>>22817
Ah, I see. Unfortunately I do not have a good Wan2.2 workflow anymore. I do not use it much because it is too slow for me, but I know the time freeze lora and the cake cutting lora are both "secret" fetish loras. Time freeze is meant for sleepy/drugged sex (and can be used for necro) and cake cutting is for gore if you turn it to low percentage and prompt correctly.

Cake cutting:
https://civitai.red/models/2168151/lx-cut-anything-its-a-cake-wan-22-i2v-a14b

>>22818
I meant the Ada Wong tentacle image. It has my original workflow in it.

If you still want a Krea 2 workflow though here is a newer one in this image (png of Yor). It's a little updated with a double-upscale for even more time wasted on generating small details LOL

to render 10s video on wan 2.2 need like 10 min on good hardware in 480p+. but because wan often generated slops, you need like 5-10x retries, and it's hours and hours.

Hi can you try to do some ass cheek removal please whenever possible. Thanks

has anyone got a good necksnap prompt? probably the one thing that ive never consistently been able to get

>>22748
Helo there, what's the prompt for this image? i've tried several variatios to no avail. Would you be so kind to share it?

>>22889
I'm afraid I didn't save the exact prompt for that one. It's surprisingly hard to replicate, I think maybe it is trained so well on boxing that it is trying to interpret everything realistically. If you describe the violence with normal boxers, it tries to make them realistic. If you say Chun Li and Mike Tyson or whoever, the gore comes out better. You have to give more detail to make it stop trying to be realistic.

This is as close as I can recreate it:

-----
Sports photo, action snapshot of a man and woman boxing. The man is throwing a heavy punch that connects with the woman's face. The woman's head explodes backwards in a splatter of gore and bone fragments and blood, crushed by the force of the punch. An eyeball flies from the impact. Half of her face is missing where the punch impacts. A splatter of deep red viscous liquid erupts from the impact of the glove.

The camera is behind the man's shoulder, following the action of the punch, with slight dutch angle adding to the feeling of forceful action, with motion blur. The background is a sports arena, blurred, with the blurred arena lights visible.

The lighting is dramatic, with high contrast dark shadows. The fighters are lit with hard white rim lighting from the stadium lights behind them.

The woman looks like chun li, wearing her blue outfit with white hair buns. The man looks like Mike Tyson, with his facial tattoo slightly visible on the side of his face.
-----

>>22866
Yeah this feels like something you'd need a lora for. It's a fight just trying to even get hands close to where they're supposed to be

-----
Photo of a woman with her body facing the viewer and her head twisted sharply and unnaturally to the side. Her hands are partially up reflexively and she has an exaggerated shocked, wide-eyed expression on her face, facing sideways.

A man in a black ski mask stands behind her, gripping her head roughly with his hands, twisting her head forcefully to break her neck. He grips and pulls her chin with one hand while his other hand pushes the side of her head, turning her head by force. There is motion blur to help convey the speed of her head whipping around. The man looks down at the woman, his eyebrows visibly scowling behind his mask.

The background is a dark rainy new york alleyway. The scene has film grain and the style resembles a TV thriller.

The man looks like Spy from Team Fortress 2, wearing his ski mask and suit. The woman looks like Miss Pauling, with her hair back in a low bun with side fringe, wearing glasses and a purple collared button down top.
-----

>>22957
Thanks a lot!, this is what i get, pretty good!

so much fun prompting in krea. realism promp tend to broken the gore thing. need use look like method to enchance gorynes

Do you have any ideas for depicting a prompt where an axe is deeply embedded in the body? If the word “axe” is included, it ends up depicting the entire axe.

>>22685
Has anyone figured out how to stop this model from forcing pubic hair?

>>23275
these loras works,
https://civitai.red/models/1838378/realistic-nudes
https://civitai.red/models/2729157/krea2-better-pussy
(no hair).

anyone know how to mask pussy area? similar as custom nodes does for faces

>>23278
try https://github.com/CliffNodes/Krea2-Multi-Character-Lora-Node-w-bounding-box lets you select areas to apply specific LORAs to. also Freefuse is insanely good at doing multiple character LORAs

>>23278
any degen specific loras out there? these are still pretty hard to work with.

krea2 is really nice

>>23455
Damn, I really need to take some time to experiment with Krea.

Has anyone managed to make something from reference image using identity edit?

>>23476
I did it,but the result is
not much well,few success.here is my workflow

>>23033
thank you for your workflow，i used make many pictures。how do you write your prompt，my prompt wrote by myself don't work well，it always can't realise what i want.but if i just change some color , clothes ,and name of character from your prompt ，it works well.here is by best work whose prompt make by myself

>>22170
can I ask prompt?

Requests?

A detailed professional digital art illustration in anime style, rendered with highly accurate anatomical precision and vibrant healthy skin tones, depicts Ayanami Rei completely naked, standing beside the viewer, leaning forward with her right hand placing a round ceramic plate onto a small table before her, her left hand resting behind her back for balance; she has short blue hair neatly framing her face, red eyes locked directly at the viewer with a neutral expression, one nipple pierced and adorned with a single nipple tag bearing the Japanese characters “同格” in clear, crisp typography; she is turned slightly toward the viewer, not seated, and gazes downward at the viewer from a Dutch angle perspective — the composition looks down at the plate and up at her, creating a dynamic, intimate interaction; the plate displays a palm-sized, centered, thin heart-shaped layer of skin and vulva spread to the sides pressed flat against it's surface, skin smooth and level with the plate’s rim and level with the clitoral hood flat against plate rendered with delicate anatomical clarity — with a very small and smooth clitoris, all illuminated by soft even lighting that accentuates texture and form without casting harsh shadows, maintaining consistent anime coloring with saturated yet natural hues; the background is out of focus, implying a romantic evening in a high-end home dining room with no animals or additional characters present, no background distractions, composition tightly framed to emphasize anatomical detail and the interaction between the character and the plate, with no extraneous elements or depth cues beyond the immediate subject, and the viewer is seated at the table being served what’s on the plate — beside the plate rests a wine glass containing three anime style eyeballs, all rendered with realistic detail and soft ambient glow, while the clitoral hood remains flush with the plate’s surface, not protruding, as if the plate itself has a heart shaped layer of skin and a pussy, and the entire scene is devoid of any extraneous objects or depth cues beyond the immediate subject.

野外求生纪录片片段，无旁白，至少四秒的延时镜头真实残酷地呈现狼群进食完整过程，视频为真实人类禁止出现机器人元素，一名日本声优女性
图片1倒在地上仰面躺着，周围围着一个数量庞大的狼群，
图片1面色惊恐发出惊慌失措的喊叫，手足无措地用手脚试图挡住周围狼群，一头狼上前叼住
图片1的脚踝向镜头外拖去，拖动中划破了
图片1裙子和袖子，随后用延时镜头形式真实残酷地呈现狼群进食细致过程，过程中可见
图片1的裙子和袖子逐渐撕裂破损，最后画面定格在图片1的衣物有撕裂状的残破，
图片1的肢体七零八落散落在场景各处，身上到处有血痕与咬痕，衣服破损下可见真实完整的专业人类解剖学结构。旁边有几只狼将前爪搭在角色身体上做出慢慢品尝的动作，
图片1脖子被叼在一只狼嘴里向外拉扯的画面上，
图片1的一只手被单独叼在一只狼嘴里

>>22674
Minimax H3 is amazing

>>23791
Indeed. Any spec. workflow or default?

>>23033
I used this workflow and it's pretty great, thank you!
One question, you have a lora titled '2C5467432VFW0Y7ENRAK07R8CD0'
Could I ask what it is and what it does?

>>23033
>>23791
ooh yeah krea + minmax is pretty good

>>23791
Yea, Minimax H3 looks like very interesting for guro-content creation

>>22466
Can someone tell me where I can find the LoRA ge_000001750.safetensors used in the work flow here can be found?

>>24466
> https://huggingface.co/bobo1666/ge/tree/main

Much thanks!

Yeah Krea is universes better than illustrious or zimage for this

>>24713
Can you do more cartoons?

>>23799
that char lora
https://civitai.com/models/2745079/bss-mila-krea2-char?modelVersionId=3087625

>>23805
wew so nice.

>>24845
wow, nice one. I've ever fantasized about girls raped and brutally murdered in the ancient and mid age times.

>>24845
😍 Very cool, they didn't even waste an arrow on a little princess, they just cut her throat 🤤🤩

>>24791
It's a little harder than realistic since the gore loras all seem trained for realism but the realistic stuff is over the line for me so if i post anything its gonna be painterly or drawn. Also the mods should probably come back in and reinforce the face that realistic ai pedo stuff is not allowed before this site gets shut down again.

>>24882
>harder
It is? I was hoping to see some ponygirls getting hurt.
>shut down
Let's do everything so this doesn't happen. Again.

>>24895
Not really my thing but maybe someone else could do the request. And yeah, we really need a culling again for a bunch of the stuff posted lately.

>>24905
I understand.

Is there any combo of prompts or loras that can get a broken neck or neck snap scene? This can do all kinds of dismemberment and gore but once the head is still on the shoulders it seems like it HAS to point in a natural direction.