/art/ - Artwork

Seems pretty useful. This grid used 8000 and got some okish results on the first try with some shit prompt. Seems to have a preferance for having breasts be exposed but other than that works pretty well. I can't make embeds cause I have low vram so I appreciate this anon.

>>36453
Same seed & prompt with 5000

We are evolving. Nice.

Ho big as the dataset you trained it on?

Was it SD 1.5?

Thanks for doing it!

>>36464
Lora - https://anonfiles.com/qc89cea7z8/hunged_girl_safetensors

Example trigger - standing girl, hunged_girl, noose, rope, hanged, asphyxiation, hands behind, <lora:hunged_girl:1>>>36464

>>36619
thanks, I managed to get it on civitai in the 30 minutes before it was taken down.
What is the base model you're using for your pictures?

>>36622
NAI for training, Grape / AbyssOrange / Mandarine for gens.

>>36622
Btw, do you managed to read comments?

>>36622
I only saw that one first commenter pearl-clutching… like "this offends me, take it down"

What a gigantic faggot

Please tell me what website it is made from. I can only make strange pictures with Novelai.

>>36628
It is not website.

It is possible using full local pr google colab installation.

Looks great OP!

>>36619
yesss a LORA now I can use it ith not anime models

>>36758
>>36759
what prompts did you use for these pics? It is much better than what I can get.

>>36787
Use the LoRA "crazy expressions" from civitai. Use it at a low weight combined with the trigger "constricted pupils". Also add crying to the prompt

>>36800
Thank you

>>36629
How do we do this,Wanting to try shit myself

>>36889
https://www.youtube.com/watch?v=VXEyhM3Djqg

Follow this to install.

then download a porn model (either an anime one or a realistic one). I must be SD 1.5
Then download the LORA in this thread

Insert a prompt, for example

Margaery Tyrell, (Game of Thrones), (hunged_girl:1.4), (noose), rope, (hanged:1.2), (asphyxiation), <lora:hunged_girl:1>, (bad_end:1.1 ), (slim), ((extremely detailed background)), (gallows:1.3), detailed facial features, big breasts, (arms behind back), (standing), (photorealistic), (naked breasts:1.25), floral pattern green dress, (mistress)

And have fun

>>36891
best thing for this LORA i that it works with UberRealisicPornMerge

is it possible to request an Impale LORA?

>>36894
Thinking about it, but need pictures for dataset. Collect them, upload archive somewhere, for example pixeldrain, and post link here.

Awesome Lora, thanks @op!

I've been trying for days to make a sword/knife insertion Lora with a dataset of around 70 images, but I can't find a good guide for this kind of Lora… (would it fall into the category of "concept training"?) ..I can only find guides for faces and styles. I am only getting shitty results with my training attempts. I suspect that I've tagged the images the wrong way. Would you please share one or two example images and tag lines? (the contents in the .txt file) And did you use regularisation images? Any info you could give in general about the settings you used to train this model would be greatly appreciated!

>>36947
No, used automated tagging by wd-tagger in lora google colab. I suggest 70 pictures is too few, hanged lora has around 400 original pictures x2 by flip aug x5 repeats per epoch.>>36947

>>36951
Thank you for the reply! Guess I'll have to get more good pictures first then…

>>36619
Could you or someone else repost the Lora? Anonfiles has unfortunately shutdown last August

>>40075
I will do one better. I compiled all the r-18g loras I have collected into a single rentry document

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

if you have more to contribute please post, i will add them.

>>40078
good work.

Your archiving work is a great idea, and I'll try to share with you the R-18G type LORA I've made/know about as much as possible.


LORA list I made:

List of R-18G type LORAs I know:

List of R-18G type LORA developed by PIXIV user deawfwaef (users/20344975):

Anyone has this model?
https://frotz.gumroad.com/l/uatfi?layout=profile

It is the only hanging SDXL model and it is behind a pay wall.

Damn

>>40078

Hidden gem on civitai: https://civitai.com/models/11496/skinless-muscle-model …might be usefull for some flaying when used in combination with inpaint. First few preview pics are a bit misleading. scroll through them to see the better examples.
____
Another amputee-lora that was on civitai for 5 seconds before getting nuked: https://files.catbox.moe/m8chiu.safetensors …original name was "amputee000.safetensors" (sry, catbox always changes filenames randomly for me)

triggerword is "amputee111".
___

Also got nuked from civit, although not directly guro but very usefull for amputee - before/after scenes: https://files.catbox.moe/n7itti.safetensors …original name was "cut line.safetensors". makes dotted cut lines (as in "cut here" - lines ) on legs and/or arms and/or the neck. very reliably working lora.

triggerwords: "cut line, body markings, arm marker, thighs marker, neck marker"

___

And another truely hidden gem that also got nuked from civit: https://files.catbox.moe/nz0hf2.safetensors

…"head_destroy.safetensors". can generate very graphic destroyed heads with exploded/missing parts of the skull and brainmatter.

Triggerwords: "head destroy, brain, blood"
…has given me the best results when used to inpaint on existing subjects, but can also be used for txt2img. (Had some promissing looking effects in combination with staravalons impalement-lora for example, where I got impaled subjects with disfigured mutilated heads plus skewered brains above the head. But I haven't figured out a good solution for the weight that doesn't overfit, yet.)

____

I've been working on a lora for vaginal sword/knife insertion/stabbing for almost a year now. I can't get the damn thing to work properly but I haven't given up just yet. Just need to create even more hq training images, but sadly I'm only working with a 1070ti… I had also planned to make a debreasting lora but since I can't even finish my 1st project I haven't even started with that.

…so if any of you more capable lora-creators needs a new project or some inspiration, I'm sure I wouldn't be the only one who'd be really happy about something for breast_amputations. Also stabbing in general. :)

>>40092
Wow. Thanks!!!

I will test them and add the repository!

I neve saw those ones.

>>40094

Oh, I forgot one triggerword for the head_destroy-lora: "headshot"

>>40094

I Hadn't used the head_destroy-lora in a while so I ran some tests again. I get the best results by inpainting (tested with a semi-realistic mix). Mask the upper head of your subject with or without the eyes. Mask additional space above and around (so there's enough space for some splatter) then set to inpaint mask only, set masked content to original, leave denoising strenght at 0.75 and cfg between 5 and 7.5. Only use the 4 triggerwords as prompt and leave lora strength at 1
so "(head destroy, brain, blood, headshot), <lora:head_destroy:1>" should be all you need.

Also tested the amputee000 lora again and I've found it to work very well in conjunction with the amputee_v5 - lora and blockweight. try these values tohether for instance + all relevant triggerwords from both loras: <lora:Amputee111:1:0.3,0,0,0,0,0,0,0,0.3,0.3,0.3,0.3,0,0,0,0,0> <lora:Amp_V5:1:0.8,0,0,0,0,0,0,0,0.8,0.8,0.8,0.8,0,0,0,0,0>

oh and just to clarify, the amputee000.safetensors really was named with the "000" when I downloaded it, even though the triggerword and the metadata have the "111". Just in case your generator doesn't automatically read the metadata correctly you may need to rename it, should you wish to try the blockweights I posted above.

>>40094

Are you still working on that rentry?

>>40291
Not really. Stuff came up.
Sorry, will try to update it.
WHy do you ask?

>>40293

Because I was wondering why the last two loras I previously uploaded weren't added to the repository (I was wondering if they just didn't work for you)
…and also because I've recently gotten a few more relevant loras that I thought should be added to your rentry.

But no pressure dude! Take our time. :)

>>40294
Ok. I will work on it today.
I Also have ac ouple PonyXL lora i need to share. Can you sahre the loras you got here? here?

>>40307

Ok, first I got this: https://files.catbox.moe/odlwiw.safetensors original filename: "amputee_v0-03.safetensors" was semi-stealthily posted on civitai as "horror anime scene - lora" with only "horror" listed as triggerword. Got nuked a day after it was uploaded. It can do nice gorey quad-amputees with open abdomens and severed limbs either laying detached from the body and/or stuffed into the pussy. Also severed heads. I haven't yet found out all it can do, you may want to check the metadata for more triggerwords, there seem to be quite a lot.
______

then I came across this: https://files.catbox.moe/mq4ojo.safetensors original filename: "lora_jugulation_v1.0.safetensors" triggerwords: "jugulation, death, 1boy, 1girl, hetero, grabbing, holding knife, bleeding" …very rigid and a bit overtrained, but usefull to create a base-image that can be refined via img2img

____

Also I've noticed some older loras where missing from the repo:

https://files.catbox.moe/yvysyt.safetensors original name: "impalement.safetensors" triggerwords: "impaled, impalement, execution" …this lycoris from staravalon is imho the best working impalement-lora. depending on your model you may want to dl the "headback" lora from civitai and/or adjust your prompt/negative prompt if you want (or don't want) the impalement-pole to come out of their mouth. sadly the original creators pixiv-acc was recently deleted, otherwise I'd link to it.

____

Sadly the same goes for this loras creator. https://files.catbox.moe/fb0syw.safetensors …original name: "ci_fu.safetensors" can generate a girl beeing stabbed with a (Katana-) sword through her belly. works better with semi realistic models, appears to be not very adaptable. main triggerword is "ci_fu". I never got great results with it but I know other people have so I'm adding this for completeness' sake.

____

And last for now is this: https://files.catbox.moe/mqepf2.safetensors "lora_dissection_corpse1.safetensors" triggerwords: "dissection_corpse, guro" …produces open abdomens. is a bit more anatomically accurate in depicting organs than the first lora in this post but also if used alone has a weird 2D-ishness to the open belly area. However at lower strength (~0.5) it improves details in conjunction with the "amputee_v0-03.safetensors" I posted above.

Hope you'll find these usefull! ^^

>>40309

P.s. All these are SD1.5 - based. I can't run xl sadly.

>>40309
>>40309
>amputee_v0-03.safetensors

This is a goldmine. Does amazing things but is messy. Strange thing that "horror" does not appear in the lora metadata.

Currnetly trying to determine which keywords are usefull and which are not.

>>40309
>impalement.safetensors
In the rentry for some reason I renamed it Impalement_civ it is the same lora.



PS Can you link you rpixiv account? So I can send you the password to modify the rentry. SO it survives even if i do not find time to upkeep it.

Added all SD 1.5 LoRAs. A couple of SDXL loras left to add.

>>40309
ci fu produced a lot of LORAS but they quality is very low. I bleive he used a very small sample size.

amputee_v0-03.safetensors

is especailly good. But is till was uanbel to put hands inside the girl

>>40317

Ai is weird! With my prefered style models the results are basicly the complete opposite compared to your 'with/without'-example. Also I would've never guessed the impalement_civ was the same lora! That example - pic in the repo looks like nothing the lora generated with my models.

I didn't even know ci_fu had more than that one lora.

>>40317

I have tried to find the right trigger for inserted limbs, too. But to me the whole lora seems to behave in quite a random fashion.
When I first downloaded the lora last saturday I just added it to an existing character-prompt that I was working on anyways and only used the triggers 'amputee' and 'gore' and immediatly got inserted limbs in every 2nd or 3rd pic. Now with a more refined prompt I don't get them either anymore. I'm sry but I can't upload an example here right now. Might be easier on pixiv.

>>40322
sent

>>40324
Yeha otputs style is quite unstable. I use a lot of negative mebeddings and thye affect the style. I go tcrazy examingin their effect matrix.

>>40318
I wonder how well that will work on 3d inpainting

>>40322
sent. In case you want to delete this message.

>>40329

yeah, thank you! And that pic is awesome! :D

>>40325

Maybe I'm misinterpreting the matrix-chart but it seems 'bad-pictures-chill-75v' has the most significant overall-effect.


>>40326

Didn't have time to run many tests but it seems that would depend on the abilities of the 3d-model you want to use. I have only a few realistic/3D basemodels and of those most can't do bloody things in general. I've only tried one model and the inpainting results were discoloured and contorted but the general concept was present. So my guess would be it should be possible but will take some experimenting.

For people here that are unable to train their SDXL lora.

Civitai allows to train models on their site. 500 buzz per geenration.

You can easily gather 500 buzz by posting comments and images or by creatign fake account and trasfering the buzz


Also a mindflayer lora came out

>>40078
Hello I'm trying to use this with ponydiffusion on forge ui and it doesn't seem to be working, I wasn't able to see the LORA unless I enabled to see all extra networks but it gives me the same results if I had no lora which is usually someone in a lot of bondage rather than noosed. What could I be doing wrong?

>>40516

Are you sure the LORA is trained for Pony diffusion? They don't have any noose related LORAs yet as far as I was aware.

For reference, 1.5 cannot be used with XL, and most XL models are trained on base XL, rather than Pony, which is such a large model that it usually has its own distinction.

>>40516

I agree with this anon >>40517

SDXL models usually do not work on pony.

Also the only SDXL hanging model is behind a paywall and was not uploaded on that rentry.


What is the excat LORA you want to use?

>>40520
Thanks I hadn't considered that, I feel really dumb now. I was trying to use the hanging models because I like hanging a lot, and I liked pony diffusion style a lot more than other models since it can do penis really well.

>>40527
>>40516
I just found a new Hanging that works for Pony XL.

you can find it here
https://www.pixiv.net/en/artworks/116669013?comment_id=173175255

Read the description. I have no idea what are the proper keywords, but it create results like picralted.

>>40530
just do not include words as hang or hanging, otherwsie it created some sort of bondage suspention.

>>40530
You're amazing! Bless you. Google was refusing to show me anything useful.