all / g / s / fur / art / ai / 3dcg / f / rp / req / p2p / dis / lit / kaki chat
[Settings] [Home] [Catalog] [Search] [Thread list] [Stats] [Reports] [Watcher] [PMs] [Admin]
[Return] [Bottom]

Posting mode: Reply
BBCode
(for deletion)
  • Allowed file types are: gif, jpg, jpeg, png, webp, webm, mp4
  • Maximum file size allowed is 25000 KB.
  • Images greater than 255 * 255 pixels will be thumbnailed.
  • 123 unique users in the last 30 minutes (including lurkers)
    First
    [1]
    		Last
    File: 1703772579609.jpg
    (197 KB, 960x1264) ▶
    202280
    No.10268 [Report] AS
    ▶

    Hello everyone,

    I'm excited to share a couple of new models I've been working on and to gather your valuable feedback. I've created a new thread to make this information easily accessible.

    1. GuRO Checkpoint Model
    https://civitai.com/models/1794928

    I have successfully trained a guro checkpoint built upon the WAI 14.0 base. This model was trained on a dataset of 6,000 images and shows promise, particularly for generating images with tags such as severed head, hang, and amputee.

    While the model is performing well, there are some known issues that I am working to improve. A comprehensive list of the tags used during training can be found here: https://rentry.co/r-18G_illustrious_tags


    Some of the tags are unconventional, but I believe they are mostly self-explanatory.

    2. Ayaswan LoRA Model
    https://civitai.com/models/1795249

    In addition to the checkpoint, I have also trained a 1GB LoRA model. This was developed using 1,500 images of Ayaswan. The tag list for this model is available at the same link: https://rentry.co/r-18G_illustrious_tags



    Important Account Update

    On a related note, my previous pixiv account was unfortunately deleted. You can now find me on my new account here: https://www.pixiv.net/en/users/101800918

    Feedback and Collaboration Welcome!

    I will be posting some sample images generated with the checkpoint and the corresponding prompts below.

    I would greatly appreciate any feedback you can provide.
    [1] Does the model work well for you?
    [2] Are there specific tags that produce better results than others?
    [3] If you create any images using these models, please consider sharing them in this thread. Seeing your creations will be incredibly helpful.

    P.S. For those interested, the checkpoint is at epoch 17 of 30, and the LoRA is at 30 of 95. If you would like an overtrained version of either , or if you are interested in the dataset used for training, please send me a message. I am also very open to suggestions on how to improve the tag lists.
    >>
    No.10269 [Report]
    ▶
    File: ComfyUI_96531_.png
    (829 KB, 768x1344) ▶
    849173
    ALL IMAGES WERE MADE IN COMFYUI.

    So just DRAG AND DROP the images to COMFYUI for the whole workflow and all positive and negative pormpts.




    asanagi,
    1girl,
    scientist, , , blonde, , elf, glasses,
    garter belt, stockings, labcoat, ,
    rope, hang, asphyxiation, noose, 1girl, solo, , breasts, open mouth, , nipples, , , , arms behind back, bound, trembling, bdsm, bondage,
    >>
    No.10270 [Report]
    ▶
    File: ComfyUI_97246_.png
    (1.19 MB, 832x1216) ▶
    1247628
    1girl, elf

    gloves,
    large breasts,
    blood, guro, longitudinal bisection, longitudinal bisection, , solo, , breasts, blush, open mouth, nipples, nude, , armpits, arms up, chain, , lactation, laying, table,
    , grabbing another's breast, pov hand,
    >>
    No.10271 [Report]
    ▶
    File: ComfyUI_97376_.png
    (770 KB, 832x1216) ▶
    789211
    asanagi,
    1girl, ,

    rope, hang, asphyxiation, ear tag, quadruple amputee, hetero, breasts, nipples, 1boy, pussy, breast grab, grabbing, lactation, , , , tongue, navel, open mouth, nude, tongue out, , bound, , solo focus, disembodied limb, white background, large breasts, simple background, restrained, bdsm, , ,
    >>
    No.10272 [Report]
    ▶
    File: ComfyUI_98631_.png
    (1.1 MB, 896x1152) ▶
    1154373
    1girl, hyuuga_hinata,

    large breasts,
    smile, open mouth, tears, crying, crying with eyes open,
    blood, eye fuck, , breasts, blush,
    1boy, cleavage, hetero, , penis, solo focus, ,
    >>
    No.10273 [Report]
    ▶
    File: ComfyUI_101444_.png
    (912 KB, 832x1216) ▶
    934096
    gblptto,
    hyuuga_hinata, naruto_(series), 1girl, black_hair, purple_eyes,
    milf,

    guro, asphyxiation, cannibalism, amputee, solo, , ahoge, breasts, nipples, bound, bdsm, bbq, bbq body, hanged, rope on neck, cooked, cooking, uncooked face, rolling eyes, ahegao, tongue out, bondage breasts, stitches on stomach,
    >>
    No.10274 [Report]
    ▶
    File: ComfyUI_101572_.png
    (1.13 MB, 1024x1024) ▶
    1180330
    ayaswan,
    hyuuga_hinata, naruto_(series), 1girl, black_hair, purple_eyes,
    milf,

    blood, guro, asphyxiation, severed head, reverse_blowjob, open mouth, multiple girls, nude, tongue, cum, blurry, cum in pussy, cumdrip, cum in mouth, ahegao, rolling eyes, death, , disembodied head, corpse, blurry background, background out of focus, grabbing naother's face,
    >>
    No.10275 [Report]
    ▶
    File: ComfyUI_101460_.png
    (1.03 MB, 832x1216) ▶
    1077428
    ayaswan,
    hyuuga_hinata, naruto_(series), 1girl, black_hair, purple_eyes,
    milf,

    large breasts,
    guro, severed head, headless, neck meat, cum in trachea, hollow_head, solo, breasts, open mouth, navel, nipples, barefoot, pussy, tongue, cum, uncensored, bound, cum on body, bdsm, bondage, facial, cum on breasts, after sex, cum in mouth, ahegao, peeing, pee, brain,
    >>
    No.10276 [Report]
    ▶
    File: ComfyUI_101477_.png
    (1.13 MB, 1024x1024) ▶
    1185207
    ,
    blood, dead, 1girl, long hair, breasts, open mouth, large breasts, black hair, 1boy, navel, holding, animal ears, nipples, weapon, hetero, nude, lying, penis, pussy, solo focus, sword, dark skin, spread legs, sex, on back, holding weapon, rabbit ears, black eyes, vaginal, dark-skinned female, completely nude, uncensored, knife, interracial, blood on weapon, chest stab,
    >>
    No.10277 [Report]
    ▶
    File: ComfyUI_101485_.png
    (839 KB, 832x1216) ▶
    859762
    Impalements do not work very well.

    but still soemtime they give funny results





    ayaswan,
    1girl,
    nipples, breasts, barefoot,
    spit_from_mouth, impaled vertical, solo, breasts, open mouth, sitting, , pussy, tongue, tongue out, , toes, uncensored, , wariza, oral, looking up, , , vaginal insertion, , , sexually suggestive, phallic symbol,
    >>
    No.10278 [Report]
    ▶
    File: ComfyUI_101503_.png
    (1.28 MB, 1024x1024) ▶
    1342206

    1girl,
    glasses, classroom background, teacher, , blonde, whore makeup,
    shirt, skirt, thighhighs, open shirt, one breast out,
    blood, cutting_throat, ,breasts, 1boy, nipples, hetero, sweat, penis, sex, sex from behind, knife, rape, injury, head grab, grabbing another's hair, hands behind back, bondage,
    >>
    No.10279 [Report]
    ▶
    File: ComfyUI_101509_.png
    (1.34 MB, 1024x1024) ▶
    1408007
    asanagi,
    1girl,
    glasses, classroom background, teacher, , blonde, whore makeup,
    shirt, skirt, thighhighs, open shirt, one breast out,
    blood, cutting_throat, ,breasts, 1boy, nipples, hetero, sweat, penis, sex, sex from behind, knife, rape, injury, head grab, grabbing another's hair, hands behind back, bondage,
    >>
    No.10280 [Report]
    ▶
    File: ComfyUI_101513_.png
    (1.04 MB, 1024x1024) ▶
    1086011
    xingzhe,
    1girl,
    glasses, classroom background, teacher, , blonde, whore makeup,
    shirt, skirt, thighhighs, open shirt, one breast out,
    blood, cutting_throat, ,breasts, 1boy, nipples, hetero, sweat, penis, sex, sex from behind, knife, rape, injury, head grab, grabbing another's hair, hands behind back, bondage,
    >>
    No.10281 [Report]
    ▶
    File: ComfyUI_101500_.png
    (1.12 MB, 832x1216) ▶
    1175773
    1girl, scientist, blonde, elf, glasses,
    garter belt, stockings, labcoat,

    hang, rope, noose, asphyxiation, arms behind back,
    large breasts,
    , solo, , breasts, blush, open mouth, nipples, nude, , , , , lactation,

    grabbing another's breast, pov hand,
    >>
    No.10282 [Report]
    ▶
    File: ComfyUI_101516_.png
    (1.17 MB, 1024x1024) ▶
    1229543
    asanagi,
    1girl, elf, queen,, mature ,

    severed head on the ground, open mouth, , long hair,

    , severed head, headless, neck meat, guro, large breasts, arms behind back, tied up, on her knees, full body,

    solo, navel,
    >>
    No.10283 [Report]
    ▶
    File: ComfyUI_101523_.png
    (1.06 MB, 832x1216) ▶
    1108904
    asanagi,
    , 1girl, large breasts,



    , nipple piercing, piercing, noose, rope, garrote, asphyxiation, purple face, 1boy, hetero, sex, leash, nude, tongue out, stomach bulge, sex from behind, tied wrists,
    >>
    No.10284 [Report]
    ▶
    File: ComfyUI_101529_.png
    (984 KB, 832x1216) ▶
    1008227
    asanagi,
    blood, guro, longitudinal bisection, 1girl, solo, long hair, breasts, blush, open mouth, bangs, large breasts, black hair, gloves, nipples, ponytail, nude, tongue, elbow gloves, tongue out, blunt bangs, armpits, arms up, chain, hime cut, lactation, disembodied limb,
    >>
    No.10285 [Report]
    ▶
    File: ComfyUI_101544_.png
    (1.01 MB, 1024x1024) ▶
    1061371
    blue eyes, light green hair, BBQ body,


    gblptto, guro, cannibalism, quadruple amputee, severed head, headless, 1girl, solo, breasts, pussy, , nipples, tongue, tongue out, open mouth, restrained, , anus, bound, bdsm, BBQ, lactation, disembodied head, disembodied head on scale, decapitation, uncooked face, cooked, cooking
    >>
    No.10286 [Report]
    ▶
    File: ComfyUI_101549_.png
    (1.01 MB, 832x1216) ▶
    1056391
    1girl, large breasts, wedding outfit, ,


    ayaswan, rope, hang, asphyxiation, noose, 1girl, solo, , , open mouth, arms behind back, bound, trembling, bdsm, bondage,
    >>
    No.10287 [Report]
    ▶
    File: ComfyUI_101551_.png
    (1.09 MB, 832x1216) ▶
    1140761
    1girl, large breasts, wedding outfit, ,


    rope, hang, asphyxiation, noose, 1girl, solo, , , open mouth, arms behind back, bound, trembling, bdsm, bondage,
    >>
    No.10288 [Report]
    ▶
    File: ComfyUI_101580_.png
    (1.18 MB, 1024x1024) ▶
    1236776
    asanagi,
    1girl,

    large breasts, completely nude, nude,
    blood, dead, breasts, open mouth, , 1boy, navel, holding, nipples, weapon, hetero, lying, penis, pussy, solo focus, sword, spread legs, sex, on back, holding weapon, , , vaginal, uncensored, knife, interracial, blood on weapon, chest stab,
    >>
    No.10289 [Report]
    ▶
    File: ComfyUI_101612_.png
    (1.03 MB, 832x1216) ▶
    1075779
    asanagi,
    tsunade_(naruto), naruto_(series), 1girl, blonde_hair, brown_eyes,



    large breasts,



    noose, rope, garrote, asphyxiation, purple face, 1boy, hetero, sex, leash, nude, tongue out, , sex from behind, tied wrists,
    >>
    No.10290 [Report]
    ▶
    Note: asanagi is not part of the trained dataset.

    I just like his style as an artist.



    P.P.S The tags that do not work very well give better resuslts with controlent of guro images
    >>
    No.10292 [Report]
    ▶
    >>10268
    thx sir! can you make the model available for generation on civitai? my desktop cannot afford to run
    >>
    No.10293 [Report]
    ▶
    very nice work!

    can you share the dataset for the v2 ayaswan so we can see how it's tagged? or maybe it'd be good to have the datasets for all of the checkpoint stuff somewhere?
    >>
    No.10295 [Report]
    ▶
    File: ComfyUI_101755_.png
    (1.4 MB, 896x1152) ▶
    1465583
    >>10292
    Hmm. Apparently civitai allows only the most popular checkpoints to be used fro on site generations. So I paid money to enable this checkpoint to be used. It still have not activated.
    Still trying to figure it out. If you have insights please share.

    THe Ayaswan LoRA is enabled for on site generations.

    As always use generic positive promp tags (best quality, masterpiecce) and negative prompts tags (bad hands, low quality etc.)


    >>10293
    I do not think sharing the dataset on this thread is a good idea. Send me a message on pixiv and I will send the dataset to you.
















    asanagi,

    1girl,
    , , scientist, , , blonde, , elf, glasses,
    garter belt, stockings, labcoat, ,



    blood, , , breasts, blush, smile, open mouth, large breasts, , 1boy, cleavage, , hetero, , penis, solo focus, tears, crying, crying with eyes open,
    brain fuck, , exposed brain,
    >>
    No.10296 [Report]
    ▶
    File: ComfyUI_101772_.png
    (1.11 MB, 896x1152) ▶
    1160798
    >>10295


    asanagi,
    1girl,
    , , scientist, , , blonde, , elf, glasses,
    garter belt, stockings, labcoat, ,

    blood, , , breasts, blush, smile, open mouth, large breasts, , 1boy, cleavage, , hetero, , penis, solo focus, tears, crying, crying with eyes open,
    hollow_head, hollow_head,
    >>
    No.10297 [Report]
    ▶
    File: ComfyUI_101793_.png
    (1.11 MB, 1024x1024) ▶
    1166998
    asanagi,
    1girl,

    large breasts,
    bite marks, bite breast, mutilated breast, breasts, open mouth, 1boy, navel, nipples, hetero, nude, penis, pussy, cum, sex, vaginal, uncensored, cum in pussy, grabbing, testicles, cum on body, breast grab, facial, empty eyes, clitoris, biting, orc,
    >>
    No.10298 [Report]
    ▶
    File: ComfyUI_101812_.png
    (1 MB, 832x1216) ▶
    1052445
    dead, hang, kurohime, noose, rope, 1boy, 3girls, 4girls, animal ears, animal print, bangs, black eyes, black hair, bound, breast grab, breasts, censored, cow ears, cow girl, cow horns, cow print, cum, elbow gloves, flat chest, gloves, grabbing, group sex, hetero, horns, large breasts, loli, long hair, motion lines, multiple girls, navel, nipples, nude, penis, print legwear, pussy, sex, small breasts, straddling, thighhighs, uncensored, vaginal,
    >>
    No.10300 [Report]
    ▶

    >>10295 (You)
    >>10292
    I submitted it for the auction.

    It should become available on Friday 0:00 AM GMT i fnothing happens.

    After that it will be available for 1 week only (unless many, many users start using it)
    >>
    No.10301 [Report]
    ▶
    File: ComfyUI_101858_.png
    (1.19 MB, 832x1216) ▶
    1245984
    asanagi,

    large breasts, mother and daughter,
    dead, hang, noose, rope, 1boy, 3girls, 4girls, animal ears, animal print, bangs, bound, breast grab, breasts, , cow ears, cow girl, cow horns, cow print, cum, elbow gloves, , gloves, grabbing, group sex, hetero, horns, motion lines, multiple girls, navel, nipples, nude, penis, print legwear, pussy, sex, , straddling, thighhighs, uncensored, vaginal,
    ,
    masterpiece,best quality,very aesthetic,highres,absurdres,
    >>
    No.10322 [Report]
    ▶
    File: ComfyUI_00068_.png
    (2.52 MB, 1544x2056) ▶
    2644774
    Tested only few concepts so far (no other loras besides detailer, just the checkpoint, either artist tags mentioned in doc, or single unrelated artist tag to check how it performs with different styles): neck meat looks solid when it works, but most of the time adds it to the places where it shouldn't be (middle of the face, shoulder, etc.), guillotine is very janky, doesn't look like one (at least i wasn't able to get the good looking one), spinal cord paired with severed head (when there is only head with no body) tends to expand neck for some reason, also i had few gens when it added ribcage to the cord too. Garrote haven't worked once for me, just never generated the device despite being paired with asphyxation (i assume it is meant for rope like strangulation, and not for garrote as execution device). Sliced off face often bleeds into disembowel, even when there is no tags related to this, same with mouth decap, but it also sometimes does the same as face off, but with neck stump cross-section on character's face, both feel super unstable. Horizontal belly slash paired with harakiri more often results in vertical cut, not horizontal. Also in general seems to be struggling with backgrounds, despite adding artist tags.
    >>
    No.10323 [Report]
    ▶
    File: ComfyUI_00069_.png
    (2.77 MB, 1544x2056) ▶
    2905513
    >>
    No.10324 [Report]
    ▶
    File: ComfyUI_00070_.png
    (2.4 MB, 1544x1808) ▶
    2513432
    Used controlnet for this one.
    >>
    No.10325 [Report]
    ▶
    File: ComfyUI_00071_.png
    (2.91 MB, 1544x2056) ▶
    3047421
    (embedding:lazypos).

    (pottsness:0.6), (cherry-gig:0.5), (artist:ningen_gokko:0.7), artist:modare:0.5, miyajima_reiji:0.5, yumeto (ym-1):0.5, (dishwasher1910:0.5), (wlop:0.4), 23i2ko:0.5, (pablo_uchida:0.5), tomatoketchup96:0.5.

    1 girl, kamisato ayaka.

    ((solo)), light blue hair, blunt bangs, pale skin:0.5, kneeling, struggling, pain, topless, hakama, harakiri, blood splatter, blood, horizontal belly slash, viscera, holding sword, tanto, looking down, blood from mouth, close-up, sweat.

    indoors, tatami, japanese room, (sanbou), from below, high contrast.



    Cut is clearly vertical, despite tags asking for horizontal.
    >>
    No.10326 [Report]
    ▶
    >>10322

    Thank you. As you have noted yourself the best results come from using controlnet on a guro image.


    Do you want an epoch 30 of the same checkpoint?

    It does guro better, but you get all the bad fingers and other early SDXL defects.
    >>
    No.10328 [Report]
    ▶
    File: ComfyUI_102151_.png
    (860 KB, 832x1216) ▶
    880990
    This model is based on illustrious. SO it works well with illustrious models.

    like Dixie/Calamity rose
    >>
    No.10329 [Report]
    ▶
    File: ComfyUI_102979_.png
    (704 KB, 1216x832) ▶
    721640
    >>10328
    >>
    No.10330 [Report]
    ▶
    File: ComfyUI_102992_.png
    (1.06 MB, 896x1152) ▶
    1110870
    >>
    No.10331 [Report]
    ▶
    File: ComfyUI_102934_.png
    (934 KB, 896x1152) ▶
    957421
    >>
    No.10332 [Report]
    ▶
    File: ComfyUI_102511_.png
    (601 KB, 832x1216) ▶
    616103
    >>
    No.10333 [Report]
    ▶
    File: ComfyUI_102529_.png
    (877 KB, 1024x1024) ▶
    898564
    >>10332
    >>
    No.10334 [Report]
    ▶
    File: ComfyUI_102609_.png
    (775 KB, 832x1216) ▶
    794447
    >>
    No.10335 [Report]
    ▶
    File: ComfyUI_103029_.png
    (870 KB, 1024x1024) ▶
    890928
    >>
    No.10337 [Report]
    ▶
    File: ComfyUI_00066_.png
    (2.64 MB, 1544x2312) ▶
    2769745
    Made quick 3d scene, and made controlnet out of it, so here is what i was able to get with more controlled spinal cord generation. Results are pretty good, but i noticed that for some reason style really fumbled, despite having quite a number of artists tags. Also what i noticed is that "empty eyes" tag bleeds into "empty eyesocket" which is definatelly a bad thing, because regular empty eyes is widelly used tag in guro gens...
    >>
    No.10338 [Report]
    ▶
    File: ComfyUI_00068_.png
    (2.53 MB, 1544x2312) ▶
    2652620
    >>
    No.10339 [Report]
    ▶
    File: ComfyUI_00067_.png
    (2.42 MB, 1544x2312) ▶
    2537605
    Eldritch horror (empty eyesocket instead of empty eyes, despite not having it in tags).
    >>
    No.10340 [Report]
    ▶
    File: ayaka_spine_canny.jpg
    (221 KB, 1024x1539) ▶
    226872
    Also canny outline, maybe someone wants to do the same thing (you can clear all unnecessary stuff like Ayaka's hair features if you want to change character).
    >>
    No.10342 [Report]
    ▶
    File: ComfyUI_103077_.png
    (1 MB, 832x1216) ▶
    1053189
    >>10337
    Yes. This exactly the kind of feedback I was looking for.
    SO in the next iteration I will replace "empty eyesocket" with soemthing else, probably just "eyesocket".

    I had a alot of problems with concept bleeding. Thank you for pointing it out.


    I need to reexamine my training data, but maybe by mistake I tagged some spinal cord protrubing from lower torso as "spinal cord" and this is why it draw ribs in the neck.

    one more thing to fix.

    This is extremely useful.

    For style you need to specify an artist. Even and artist not on the lsit that WAI/illustrious knows.

    Fo example asanagi is not part of the trained dataset but it really help my pics.

    I do not know why style is so shit if I specifiy nothing.
    >>
    No.10343 [Report]
    ▶
    File: ComfyUI_103090_.png
    (1 MB, 1024x1024) ▶
    1050603
    >>10342
    >>
    No.10345 [Report]
    ▶
    File: ComfyUI_103146_.png
    (784 KB, 1024x1024) ▶
    803330
    >>10343
    Reiq used as an artist tag seems to produce clean results.
    >>
    No.10352 [Report]
    ▶
    File: ComfyUI_103539_.png
    (1.22 MB, 832x1216) ▶
    1283098
    Wedding dresses are sexy
    >>
    No.10355 [Report]
    ▶
    Wonderful model.
    It generate naked female body very often, appropriate clothing tags (or "cosplay" tags) will solve the problem.
    >>
    No.10356 [Report]
    ▶
    I’m currently focusing on the Ayaswan_IL2 LoRA. From my brief tests, both image quality and generation stability have improved markedly over the previous version. Below are a few issues I noticed along with some suggestions:

    Transversal / Longitudinal bisection
    This tag pair still performs poorly—I haven’t been able to generate a true bisection with either variant. Because “transversal” and “longitudinal” are separate tokens, the model seems to merge them into a single generic idea of “bisection,” producing only the common features of both (e.g., disembowelment and exposed guts) instead of the specific cut. For similar concepts that differ only in orientation, try combining the words with an underscore—e.g., transversal_bisection and longitudinal_bisection. Treating each as a single token should help the model distinguish them more reliably. Similar issues was also reported by others for horizontal/vertical belly slash.

    Neck meat
    I assume this is intended to represent a neck stump? The issue here is that the base model already has a strong prior understanding of “meat,” so it tends to render the cross-section like a slice of fresh steak. This visual often bleeds into other amputation areas in the image if there are any.

    Foot / Leg / Arm Insertion
    I haven’t been able to generate correct limb insertions using text-to-image; the model tends to insert unrelated objects instead. Based on my past experience with rbqinori_v8, this type of concept generally performs much better with inpainting rather than direct generation.
    >>
    No.10357 [Report]
    ▶
    File: transversal bisection.png
    (3.49 MB, 1472x1472) ▶
    3655801
    >>10356
    1girl,solo,transversal bisection,lying on bed, nurse, guro, blood, embedding:lazypos,

    Issue: no bisection, only disemboweling (longitudinal as well)
    >>
    No.10409 [Report]
    ▶
    File: ComfyUI_105206_.png
    (1.11 MB, 1024x1024) ▶
    1163136
    >>10356
    Thank you!!!

    hmmm. Another user told me that SDXL turns all "_" into spaces. I have not tested it yet.

    But I agree they should be rolled together. I will probably use a normal dash "-".

    yes, neck meat was for the neck stump. I didn't want to use "neck stump" cause AI sometime draw a tree stump when i ask them for it. I will probably use "neck-meat" next run.

    In general I need to improve my tags.

    Some of them are too specific , and most of them are not specific enough.

    Also I will need to create synthetic data for concepts the model struggles with.
    And most importnat the concepts model struggles with and people want to use. Like the leg/arm insertion.

    So I will do a bunch of synthetic data for Transversal / Longitudinal bisection and for Foot / Leg / Arm Insertion.
    Maybe sliced off-face

    if you have sem more tags to imrpove for th enext version feel free to write them.


    And thank you for testing the model.
    >>
    No.10411 [Report]
    ▶
    This is great! Do you mind if I upload it to Tensor Art? Or maybe you could do it yourself?
    >>
    No.10412 [Report]
    ▶
    >>10411
    I do not have an account there yet. Feel free to upload it anywhere.
    Add some instructions.
    >>
    No.10413 [Report]
    ▶
    File: 00077-3722115387.png
    (2.95 MB, 1440x2160) ▶
    3092884
    I'm struggling to get good results out of it in forge. It really wants to be low res and unstable.
    >>
    No.10414 [Report]
    ▶
    >>10413
    the model is quite unstable. I know. Hopefully I will manage to fix it in version 2.

    So yes, it is not your fault.

    It works best with a controlent,
    >>
    No.10416 [Report]
    ▶
    Thanks for adding this! These are all amazing and I downloaded the model from civitai. I'm definitely going to try and use it myself but keep posting! Especially Dixie/Calamity Rose from Angel Corps!!
    >>
    No.10417 [Report]
    ▶
    File: ComfyUI_106180_.png
    (1.14 MB, 1024x1024) ▶
    1199639
    I managed to make the checkpoint model active for generation for 1 week.
    on civitai.

    https://civitai.com/models/1794928

    reminder that is quite unstable and I am collecting feedback on which exact things people want fixing.
    >>
    No.10418 [Report]
    ▶
    >>10413
    >>10413
    I love this execution of Tsunade
    >>
    No.10420 [Report]
    ▶
    Thank u very much for the great job!I cant imagine it works on civitai today!
    I think the instability cames from ayaswan style since his artwork splatter blood everywhere. While using ayaswan lora alone, the result is also unstable. Prompting with artist name and specific character name slightly improve the quality. After a dozen of failures, I prefer to use the basic WAI_NSFW with your loras. And turn the weight of ayaswan below 0.5 and it may return creatives.
    >>
    No.10421 [Report]
    ▶
    File: JZF9ZS9KW4PNWDG0G4DSCGXA50.jpeg
    (209 KB, 832x1216) ▶
    215009
    I want to share my quick&cheap workflow on civitai. Your loras work very well with draft mode of WAI_NSFW and the draft mode often get better quality than non-draft. Using 0.3~0.7 your loras that make the generate know guro tags, and mix with another random character lora with the weight 0.9+ that ensuring global quality and style. If good results come out, then start another img2img or face-fixing workflow. The CFG Scale of img2img workflow should be less than 3, and steps should be less than 15. This workflow both keeps guro tags and the high quality of WAI_NSFW. Besides WAI,another 3 modules also work well with your loras. iLustMix for realistics, Prefectious XL NSFW for strong women , JANKU v4 NSFW did better than WAI on high light and color. The attachment was generated with draft mode, mixing with some of your loras with different weights.
    >>
    No.10422 [Report]
    ▶
    File: S6C3MWFBQW0D2G78MZRNXZ89K0.jpeg
    (170 KB, 832x1216) ▶
    174245
    Another example, draft mode, low weight for guro loras, and 0.95 weight for the artist style JuanGotoh.I knew JuanGotoh also draw guros, your guro loras inspires it out
    >>
    No.10423 [Report]
    ▶
    File: K8BF6XPJ36W481CR300DH53YF0.jpeg
    (189 KB, 832x1216) ▶
    194154
    another, iLustmix module, low-weighted guro loras, and a random-found character lora.
    >>
    No.10424 [Report]
    ▶
    File: PJFEN6D3SCV13R27XX1QF5PWD0.jpeg
    (274 KB, 832x1216) ▶
    281432
    The module of JANKU v4 NSFW, high light&saturation. With very low weight (0.1~0.2) of guro loras. This one is not a draft, but 14 steps only. Scat was limited by civitai so I use alternatives.
    >>
    No.10425 [Report]
    ▶
    It is my personal opinion. Training a checkpoint might not be a good idea. While I generate pictures, i don't want to mix concepts up. I don't want every harakiri girl with a noose on neck, neither want every hanged girl dissected. Specific loras are more flex than a compilative checkpoint. The checkpoint's quality is not as good as the base module WAI, and I can't use iLustmix or other modules to draw other styles. The last and the fatal one, the checkpoint affact face too much. That side-effect requires high weight of other character loras to override that may cause distortion.
    >>
    No.10426 [Report]
    ▶
    I think the problem of Ayaswan style is, it always want to tear the girl up whereas other modules&loras basiclly build whole girls. I haven't made a test, but Ayaswan lora might perform better with controlnet or img2img workflow with high denoise level.
    >>
    No.10444 [Report]
    ▶
    acturally I quite like the idea of a checkpoint for those r-18g content.whith this I can easily put different guro content in one image.

    Is there any method to make the head also show the color of rosted meat? I found very difficult to make head rosted weither in this checkpoint or bbq lora
    >>
    No.10448 [Report]
    ▶
    >>10444
    Use regional prompt for head add roasted tags only to the head part.
    >>
    No.10451 [Report]
    ▶
    in YB's lora there is head trauma prompt, is that included in this checkpoint?
    >>
    No.10453 [Report]
    ▶
    >>10423
    >>10421
    Wow. Thank you for trying the model.

    I am quite curioous about the rest of pictures done with a realism LoRA

    If you have more could you share them on Session?

    0598027fd89ed8a444b740f8679e60bf2e273dfa1441a5984c2a59f28183831628


    I agree that individual lora are better, the checkpoint makes sense if you want to combine concept people rarely combine.

    I think I need to create more synthethic data for the next batch.

    Thank you.


    >>10444
    In theory the tag "BBQ body" makes the body cooked and the head no. And "full BBQ" makes all of the body cooked. But it is not working correctly

    I think i need better tags, like "cooked body" and "cooked head"


    >>10426
    >controlnet
    Yes, Both perform much better in contorlent.

    >>10451
    What is YB?

    Hollow_head tag should make an empty hole in the head appear
    >>
    No.10454 [Report]
    ▶
    >>10453
    sorry but i don't know what the "session" means exactly.
    I only made few realisms,
    and i don't think it is a good idea to post too realistic pic on GuroChan.
    Just try the module of iLustmix or PerfectDeliberate. Some other module also work.
    With your loras, mr_kurz:0.6,ayaswan:0.2.
    Other character lora is not very necessary.
    The most important thing is, Using draft mode. If the draft of the module perform terrible, try CFG less than 3, step less than 14.(This may be counterintuitive, but actually both keep the quality of module and the action from lora).
    Besides your tags,add guro tags found from Danbooru, such as "rope_around_neck" "intestines" "disembowelment", believe me that works and can be scaled by your loras."detailed face" or "detailed katana" are always useful.
    Some results have good pose but bad quality, however they are drafts.Then start a face_fixing or img2img workflow to improve the quality.
    While img2img, turn the weight of mr_kurz less than 0.2,10 steps only,denoise less than 0.35. Otherwise it affact the face a lot.
    >>
    No.10455 [Report]
    ▶
    File: 6N4DBX19MX0WH5QPV6R5BRTHE0.jpeg
    (149 KB, 832x1216) ▶
    153201
    just sharing some other works made with your loras these days.
    This one using the module of PerfectDeliberate and m0zg:0.35, draft mode.
    I think the original picture is too realistic for gurochan that i slightly "cartoonlize" it with img2img.
    >>
    No.10456 [Report]
    ▶
    >>10453
    if you search yb_il you could see the artstyle and examples of yb's style,which include head trauma, a kind of smashed head. And in your ARTSIT TAGS there is yb so I believe you have included his contents in this style.
    >>
    No.10457 [Report]
    ▶
    some tips,
    plz put mr_kurz or m0zg to NEGATIVE prompting. I know that is also counterintuitive.
    That won't disable guro tags but reduce the affact to faces.
    The tags of monochrome&greyscale shoud also be placed at negatives.
    >>
    No.10458 [Report]
    ▶
    File: MV8XHC0Q82RDH0YSHG9QBZ8M70.jpeg
    (240 KB, 1216x832) ▶
    246343
    WAI_NSFW,m0zg:0.3,mr_kurz:0.1 ,draft,
    face-fixing after generated.
    >>
    No.10459 [Report]
    ▶
    File: SY758S56586VQXWWWPA0701SK0.jpeg
    (220 KB, 1216x832) ▶
    226233
    WAI_NSFW,m0zg:0.3,mr_kurz:0.7 ,draft,
    face-fixing after generated.
    >>
    No.10460 [Report]
    ▶
    File: YDKYZMBCQQ98TWBZ3CAHFTA080.jpeg
    (163 KB, 832x1216) ▶
    167399
    WAI_NSFW,mr_kurz:0.3,a random-found sakura lora:0.9,draft.
    >>
    No.10461 [Report]
    ▶
    File: FY9V1G98ER7S902F4TQ7Y9CCA0.jpeg
    (84 KB, 720x824) ▶
    86027
    img2img. Source pic from 吉光骨食.
    JANKU v4 NSFW
    >>
    No.10463 [Report]
    ▶
    File: Y427B2BESFSDKD5BYJ41E2KXG0_edit_1319853(…).jpg
    (220 KB, 652x992) ▶
    225421
    Img2img, WAI_NSFW, the source pic is a 3DCG from aya posted on pixiv.CFG:0.25,denoise:0.3,mr_kurz:0.2,an oiled-up lora:1.0.

    With the workflow, I also made img2img with the source pic from ACB,aya(3dcg),pokkuti(muscular),and others.
    I think the quality is incredible.
    thx for the great loras again!

    The lora of mr_kurz is compatible with almost all the situations.
    It still has some issues, such as affacting the face and pose much, always replace fire with lava, links 2 girls with intestines. But all of them are minor.
    >>
    No.10517 [Report]
    ▶
    File: ComfyUI_109418_.png
    (1.41 MB, 896x1152) ▶
    1475181
    >>10463
    I love this one. beautiful.

    >>10454
    >sorry but i don't know what the "session" means exactly.
    Session is an app like telegram and whatsapp. But with better anonymity and protections.


    the code i sent is to contact me.

    I really apprecaite your feedback and how you tested the model. Thank you very much!


    P.S. I'm trying the model with various LoRA and they do combine nicely.
    >>
    No.10519 [Report]
    ▶
    File: S9GGZSV2NBC2C0H35PW7HNK120_edit_8309588(…).jpg
    (284 KB, 832x1216) ▶
    291805
    >>10517
    I tried some lora these days, and here is the feedback.
    While other loras describe the body well, (such as bodysuit,skins,or muscle),then your checkpoint won't tear girls up, the body is good-looking.
    But the face is the problem. Your checkpoint contains many broken or low-quality faces and line-arts.Dolcett faces or m0zg faces are quite difficult to be combined with other colorlized artworks, and ayaswan always make faces crazy, let eyes watching different sides.
    By using specific character lora, especially characters that not included in your checkpoint(e.g. Dixie) with high weight could override the face and improves the quality.
    But,high weight of loras made img2img almost impossible. Then I need turn back to WAI or others, and start another img2img for better quality.
    The attachment using the G-modile, a muscular lora and character lora, 400 buzz coin and 50+ failures. The quality is still far away behind WAI but acceptable.
    BTW, it isn't convenient for me to register a Session.
    I will continue pay attention on this thread.
    >>
    No.10520 [Report]
    ▶
    File: 268C5BNP7YYJWMS2WJWHKDM440.jpeg
    (250 KB, 1024x1024) ▶
    256528
    random pics that generated with low-weight of your loras.

    WAI, draft, using the artstyle loras of pokkuti, mid-weight of a realistic muscle lora, and oiled up lora.
    >>
    No.10521 [Report]
    ▶
    File: CR5EN38G8Q1V919GNBBZCAPNJ0.jpeg
    (158 KB, 504x888) ▶
    162407
    I hope this one doesn't offend the artist I respect. If it does,then tell me to delete it.
    Everyone here know the classical source from JuanGotoh.
    I fill line-art with simple color blocks by PS, and img2img with JuanGotoh artstyle lora weights 1.0.
    Your loras are necessary, even it is of very low-weight. Otherwise WAI directs the knife outside.
    >>
    No.10522 [Report]
    ▶
    File: 161CAYHYHTT0XHCKA7MV8EMNX0.jpeg
    (97 KB, 560x800) ▶
    99381
    I once tried to cooperate with ACB the artist to make games. Long time no contract.
    The img2img quality is much higher than expected.
    >>
    No.10523 [Report]
    ▶
    File: 61B6N7WA4AAAHZV8145BSTKA90.jpeg
    (304 KB, 832x1216) ▶
    311733
    A commision from my friend.
    >>
    No.10525 [Report]
    ▶
    File: PRHPJYFB71MEBVF9QZWMGY9C60.jpeg
    (193 KB, 832x1216) ▶
    197687
    another commision,draft quality
    >>
    No.10529 [Report]
    ▶
    I have to say i found two kinda illogical problems during three consecutive days when i was using some different IL loras with you checkpoint.
    Firstly for the "tranverse bisection"and "longitudinal bisection" two tags,the former works very well as long as i give it a strong weight,but longitudinal bisection seems to produce the same result as the "tranversal",namely only creates a horizontal cut,not a vertical one.
    And there still exists a situation that when i use tags such as "severed head" or"disembodied head",two or more than two heads are produced which is very ambivalent with a single torso.And so do the limbs.
    I also found that most time i want to make a preson and his head seperate,the head always appears quite closely to the neck of the person.Luckily i get result i want,the seperated head is mostly short hair,though the character obviously have a long hair.
    that is the feedback i want to provide. btw i am not a native english speaker.so some phrases might look strange.
    >>
    No.10534 [Report]
    ▶
    I wonder if training several distinct concept loras with data from various artists would provide better accuracy than a compiled checkpoint
    >>
    No.10539 [Report]
    ▶
    File: ComfyUI_109759_.png
    (768 KB, 896x1152) ▶
    786712
    >>10534
    Many people I spoke to believe that yes, it would be better.


    Right Now I am generating synthethic data for the next attempt. Will probably train both a LoRA and a Checkpoint on that data.
    >>
    No.10562 [Report]
    ▶
    Thank you for your work
    >>
    No.10579 [Report]
    ▶
    File: ComfyUI_110426_.png
    (736 KB, 896x1152) ▶
    753788
    >>10356
    >>10357
    I just checked my dataset.

    And for some reason I mixed up the two concepts.
    Longitudinal bisection was applied to transversal images and vice versa.


    So if you want a vertical cut you must use.
    Still it wokrs only with very strong controlNET.

    liek picrelated. I never managed to make it work as a standalone.


    might train a specific LORA
    >>
    No.10587 [Report]
    ▶
    >>10579
    what is the effect of taxidermy? I used that, and it triggered something, but the the result is very distorted picture.
    180 head twist and broken neck seemed trigered nothing on picture
    >>
    No.10594 [Report]
    ▶
    File: 00039-2369729515.png
    (1.26 MB, 1048x1048) ▶
    1322607
    it will be better if that model make more various styles,
    such as exquisite anime screenshot-like-images.
    >>
    No.10598 [Report]
    ▶
    File: 00025-6113210.png
    (1.55 MB, 1152x1328) ▶
    1625757
    If more various Lora's reflection rate improves, there will be more reasons to use this model.
    I understand that it is difficult, though.
    >>
    No.10607 [Report]
    ▶
    File: 00005-1038607730.png
    (1.59 MB, 960x1440) ▶
    1666461
    People are being way too harsh to this checkpoint. It's unstable and doesnt interact with some loras well, but punishment isnt much different. I've gotten some good results out of it now that I know how to use it, and honestly most guro loras barely work in the first place so the checkpoint helps.
    >>
    No.10610 [Report]
    ▶
    File: ComfyUI_110627_.png
    (1.09 MB, 896x1152) ▶
    1140875
    I made a vertical bisection LoRA
    https://mega.nz/file/YN5iFKZb#3STzLuztjqMjqAFw4tnLcrNIDg19qPUSw6DQTuDwroE


    tag: bisection
    Strength: 1-1.2

    for Illustrious
    >>
    No.10611 [Report]
    ▶
    File: ComfyUI_110705_.png
    (1.14 MB, 1024x1024) ▶
    1193961
    >>10610
    >>
    No.10635 [Report]
    ▶
    can I ask what happen to the model in civitai somehow I can use it direct on civitai any more
    >>
    No.10640 [Report]
    ▶
    >>10635
    Probably got banned or deleted, that happens a lot to guro stuff.
    >>
    No.10641 [Report]
    ▶
    File: ComfyUI_111469_.png
    (1.02 MB, 1024x1024) ▶
    1070479
    >>10635
    >>10640
    Civitai allows only 200 models at a time to be used for creation.

    last week i spent real money to put my model in top 200 to allow users to use it.


    THe week ended today. Since it was not popupalr enought to garantee a permanent spot it got shifted.

    You need to build around 10 USD to get a model up for 1 weeThis was mentioned in the one of th eposts above
    >>
    No.10646 [Report]
    ▶
    File: 00026-1040733384.png
    (1.31 MB, 1024x1536) ▶
    1373168
    >>
    No.10647 [Report]
    ▶
    File: 34343343433.jpg
    (716 KB, 1024x1024) ▶
    733629
    img2img
    >>
    No.10650 [Report]
    ▶
    File: 3HV6FB7VCXXAD97TGN709WRJG0.jpg
    (373 KB, 832x1216) ▶
    382691
    great model sir!hope to bring it back(for generation on civitai)
    >>
    No.10662 [Report]
    ▶
    >>10650
    >hope to bring it back(for generation on civitai)
    unless someone is willing to shell 10 USD that is not happening.
    >>
    No.10664 [Report]
    ▶
    >>10662
    Can this be done via civitai buzz?
    >>
    No.10666 [Report]
    ▶
    >>10662
    How long does that keep it up for?
    >>
    No.10667 [Report]
    ▶
    >>10666
    I have send 10 buzz to the checkpoint so I think it got enabled sooon
    >>
    No.10669 [Report]
    ▶
    >>10666
    1 week. You need to win the auction (roughly 10 USD) every week
    >>10664
    yes, the auction is via civitai buzz

    >>10667
    yes. Now that I have 10 buzz we are super close to the require 10 000 buzz required.
    >>
    No.10670 [Report]
    ▶
    >>10669
    Can I ask one thing how the heck some of models in tops with only one buzz
    >>
    No.10692 [Report]
    ▶
    >>10669
    stonks
    >>
    No.10694 [Report]
    ▶
    >>10670
    They probably pay the money. I dont get why people would use civit to make guro pics anyway, thats leaving one hell of a paper trail.
    >>
    No.10732 [Report]
    ▶
    >>10670
    the buzz displayed near the model is how much money was donated to the model.

    The buzz paid to tauction is not shown there.

    So it is irrelevant.
    >>
    No.10766 [Report]
    ▶
    File: ComfyUI_temp_gsnlz_00125_.png
    (2.23 MB, 1024x1280) ▶
    2340672
    >>10356
    >>10409

    Continued testing of the Ayaswan LoRA:

    Penis from mouth:
    Very low success rate (~5%). It usually generates a penis entering the mouth, even when using strong negative prompts to exclude "oral" and "fellatio".

    Half head:
    I expected the top half of the head to be missing, but the output mostly shows general injuries around the head or neck area instead.

    Mouth decap:
    I'm not entirely sure what this tag is supposed to trigger. There are two main outcomes:

    1. The character’s mouth is stretched open to an exaggerated, almost comical extent.

    2. The character’s entire face is missing and replaced with a large, gaping maw.

    prompt for the attached image:

    pos: 1girl, mouth decap,kushima kamome, simple background,blood,guro, embedding:lazympos,

    neg: embedding:lazyhand,embedding:lazyneg,
    >>
    No.10767 [Report]
    ▶
    File: ComfyUI_temp_gsnlz_00127_.png
    (1.58 MB, 1024x1280) ▶
    1661702
    >>10766
    I think this might be the intended effect for "mouth decap"?
    Got it after about 20 more gens using the same prompt setup.
    However, I couldn't reproduce it when I added "severed head" to the prompt; it always ends up generating a cleanly severed head with the neck still attached.
    >>
    No.10780 [Report]
    ▶
    neck fuck and cum in trachea,which one corresponds to cuming in the neck from the body,and which one to cuming in the neck from the head.And also what does trachea tongue means?
    >>
    No.10941 [Report]
    ▶
    File: F6CKZ9C1Z2TG2J243VHGV69YZ0_edit_8657065(…).jpg
    (707 KB, 2144x2752) ▶
    724906
    workflow:img2img(repeat 3 times)
    check point:Unholy Desire Mix - Sinister Aesthetic (Illustrious)
    CGF：0.3
    denoise:0.5~0.55
    Mrik's amazing loras only:mr_kurz/ayaswan:0.15~0.5(changed between workflows)
    source pic from:吉光骨喰(authorized!)
    >>
    No.10942 [Report]
    ▶
    File: QBT13KR31MWWVH7FTKYRJ0QEQ0.jpeg
    (523 KB, 2000x2168) ▶
    535648
    2
    same workflows
    >>
    No.10943 [Report]
    ▶
    File: VPD7KPSZFFH8PFNCB3AZM7R2B0.jpeg
    (201 KB, 832x1216) ▶
    206662
    3,
    similar workflows,
    denoise 0.75,repeat 3 times.
    keywords in addition such as "oni,scars,horns"
    >>
    No.10988 [Report]
    ▶
    >>10943
    good，more please
    >>
    No.11297 [Report]
    ▶
    File: ComfyUI_113207_.png
    (1.31 MB, 1024x1024) ▶
    1370424
    >>10766
    Half head and Mouth decap
    I used it is to tag both for only the upper part of the head and the bottom part where the head is missing, bad idea in retrospect.

    >>>10780
    cum in trachea is th eone with cum in the neck of the body. not relatd to the head
    >>
    No.11298 [Report]
    ▶
    File: ComfyUI_113256_.png
    (1.16 MB, 896x1152) ▶
    1213104
    >>
    No.11299 [Report]
    ▶
    File: ComfyUI_112986_.png
    (1.16 MB, 1024x1024) ▶
    1211477
    >>
    No.11300 [Report]
    ▶
    File: ComfyUI_112815_.png
    (1.01 MB, 1024x1024) ▶
    1060225
    >>
    No.11301 [Report]
    ▶
    File: ComfyUI_113278_.png
    (1.22 MB, 832x1216) ▶
    1282288
    >>11300
    >>
    No.11302 [Report]
    ▶
    >>10767
    >>10766
    any chacne you have some pics of the mouth decap, so I can add them to the dataset next time I train it?
    >>
    No.11424 [Report]
    ▶
    >>11301

    Damn! How did u make this? Where? Whats the prompt? Good work!
    >>
    No.11464 [Report]
    ▶
    >>11424
    I used the checkpoint I shared above, with tags:
    hang, asphyxiation, rope, noose,

    and i added a kingbang LoRA from civitai
    THere are severl, I used this one
    https://civitai.com/models/1433478/kingbang-art-style-ponyilsdsdxl
    >>
    No.12449 [Report]
    ▶
    great lora! great job!
    >>
    No.12497 [Report]
    ▶
    File: 00047-1760324089.png
    (568 KB, 600x800) ▶
    582558
    Awesome model, thanks for sharing! Here are a few I've made using it.
    >>
    No.12498 [Report]
    ▶
    File: 00184-681845158.png
    (598 KB, 600x800) ▶
    613105
    >>12497
    >>
    No.12499 [Report]
    ▶
    File: 00197-1008966214.png
    (638 KB, 600x800) ▶
    653724
    >>12498
    >>
    No.12500 [Report]
    ▶
    File: 00206-2418962781.png
    (589 KB, 600x800) ▶
    604058
    >>12499
    >>
    No.12501 [Report]
    ▶
    File: 00220-987133564.png
    (586 KB, 600x800) ▶
    600309
    >>12500

    This one is my favorite.
    >>
    No.12502 [Report]
    ▶
    File: 00226-3235697609.png
    (556 KB, 600x800) ▶
    570281
    >>
    No.12503 [Report]
    ▶
    File: 00235-2061970962.png
    (665 KB, 600x800) ▶
    681413
    >>12502
    >>12501
    >>
    No.12504 [Report]
    ▶
    File: 00014-2768806777.png
    (802 KB, 720x960) ▶
    821796
    >>12503
    >>
    No.12505 [Report]
    ▶
    File: 00025-2235963737.png
    (812 KB, 720x960) ▶
    832294
    >>12504
    >>
    No.12506 [Report]
    ▶
    File: 00034-2399883065.png
    (770 KB, 720x960) ▶
    788814
    >>12505
    >>
    No.12507 [Report]
    ▶
    File: 00037-991023672.png
    (739 KB, 720x960) ▶
    757401
    >>12506
    >>
    No.12519 [Report]
    ▶
    >>12507
    >>12506
    >>12505


    They look great. I'm happy someoen is using this model.


    Also I am sorry for not making a version 2. was too busy, barely touched AI>
    >>
    No.12550 [Report]
    ▶
    >>12519
    https://www.pixiv.net/users/71952036
    this is my pixiv account where i put my ai works using your checkpoint,you can have a look on it.
    >>
    No.12554 [Report]
    ▶
    File: ComfyUI_temp_jeibg_00117_.png
    (2.21 MB, 1024x1280) ▶
    2313182
    >>11302
    I generated a dataset using the one picture I post as controlnet. I tried to diversify the characters so as to not cause overfitting in training. But the camera angle is always the same since it's from the same conrolnet, you will need to flip the images or even better using Qwen image edit to do some angle variations as training materials.

    https://drive.google.com/file/d/1257K-_Ywdf9QiU5UXKlNTOtfMnuYybf0/view?usp=sharing

    the password for the file is gurochan.
    >>
    No.12555 [Report]
    ▶
    >>12554
    I also suggest you change the trigger word from mouth decap to something else next time. Since I discovered that I have much higher success rate for generating this concept when I add "mouth" to the negative prompt. Meaning that "mouth" word in mouth decap is triggering the model to draw a complete mouth, which is fighting against the idea of "mouth decap"(removal of the lower jaw).
    >>
    No.12596 [Report]
    ▶
    File: ComfyUI_115991_.png
    (1.03 MB, 832x1216) ▶
    1084021
    >>12554
    >>12555

    Thanks. Now i onl need to find the time to do another round of training.
    >>
    No.12608 [Report]
    ▶
    >>12596

    sir, Where can I get the checkpoint "Big_r-18G_1-000017"?
    >>
    No.12624 [Report]
    ▶
    File: ComfyUI_117390_.png
    (1.18 MB, 1216x832) ▶
    1240145
    >>12608
    it is this one
    https://civitai.com/models/1794928
    >>
    No.12641 [Report]
    ▶
    >>12624

    ty
    >>
    No.12671 [Report]
    ▶
    File: ComfyUI_119900_.png
    (1.17 MB, 1024x1024) ▶
    1231602
    I tried some decaps. some are cool But we need a V2 of the model
    >>
    No.13016 [Report]
    ▶
    File: ComfyUI_122001_.png
    (978 KB, 832x1216) ▶
    1002306
    Ok. I am actually at a stage where I make the V2 of my Checkpoint.


    I struggle to find good pics of ear fucking.

    The best I can generate for reference is this.

    Anyone have examples of ear fucking that are not bad nor have animal ears?
    >>
    No.13021 [Report]
    ▶
    Also any good debreasting lORA or pics?

    Posisbly debreasting LoRA.
    >>
    No.13033 [Report]
    ▶
    good to know you are making an update. in the new version, please include cooked face, and also the bruise on neck or rope marks caused by hanging/strangling
    >>
    No.13055 [Report]
    ▶
    one thing I noticed in the v1 is that penis_insertion or penis_Penetration can easilly generate similiar result with penis through stomach, maybe modify training prompt a bit since penis_insertion is really needed in layered prompts
    >>
    No.13139 [Report]
    ▶
    File: ComfyUI_127501_.png
    (1.19 MB, 896x1152) ▶
    1244134
    >>
    No.13143 [Report]
    ▶
    File: ComfyUI_125655_.png
    (1.2 MB, 1024x1024) ▶
    1256248
    >>
    No.13145 [Report]
    ▶
    File: ComfyUI_125741_.png
    (1.4 MB, 1024x1024) ▶
    1464512
    >>13143
    >>
    No.13146 [Report]
    ▶
    File: ComfyUI_126922_.png
    (1.03 MB, 832x1216) ▶
    1079692
    >>13145
    >>
    No.13147 [Report]
    ▶
    File: ComfyUI_126920_.png
    (1.08 MB, 832x1216) ▶
    1127542
    >>13146
    >>
    No.13148 [Report]
    ▶
    File: ComfyUI_127156_.png
    (1.2 MB, 832x1216) ▶
    1256033
    >>13147
    >>
    No.13149 [Report]
    ▶
    File: ComfyUI_127366_.png
    (1.14 MB, 1024x1024) ▶
    1198024
    >>13147
    >>13146
    having trouble making a cooked head.
    >>
    No.13150 [Report]
    ▶
    File: ComfyUI_128388_.png
    (1.45 MB, 832x1216) ▶
    1520984
    >>
    No.13151 [Report]
    ▶
    File: ComfyUI_118594_.png
    (1.22 MB, 768x1344) ▶
    1282172
    >>13150
    >>
    No.13152 [Report]
    ▶
    File: ComfyUI_118615_.png
    (672 KB, 832x1216) ▶
    688427
    >>
    No.13153 [Report]
    ▶
    File: ComfyUI_118663_.png
    (519 KB, 768x1344) ▶
    532254
    >>
    No.13154 [Report]
    ▶
    File: ComfyUI_118761_.png
    (1.04 MB, 768x1344) ▶
    1095030
    last dolcett
    >>
    No.13155 [Report]
    ▶
    File: ComfyUI_129208_.png
    (608 KB, 1024x1024) ▶
    622842
    >>
    No.13156 [Report]
    ▶
    >>13152
    Does the checkpoint also have a horizontal spitroast option?
    >>
    No.13157 [Report]
    ▶
    File: ComfyUI_129396_.png
    (1.05 MB, 1536x640) ▶
    1102425
    >>13156
    it has it. But it does not work very well.
    verson 2 should fix it.
    right now only 1 out of 10 horizontal spitroasts come out ok.
    >>
    No.13158 [Report]
    ▶
    >>13155
    love the facial expression on this one
    >>
    No.13179 [Report]
    ▶
    >>10409
    I've been really struggling with a bunch of severed head related pics. I really love it when the neck stump is accurate and detailed and "neck meat" seems to be the best tag to use, but doesn't always have much of an effect, or breaks things entirely.

    I'm also really struggling to get that open empty skull look you have in this pic. I want to make pics where the girl's head is being used as a bowl, full of cum/piss, or smashed brains, or even something else, but I can't consistently get that look with anything I try. "Exposed brain" gets me a general open head look, but I don't want the brain there. "brain dump" had odd and interesting results, but mostly was too much like "exposed brain". I've tried long and short descriptions, and exhausted so many keyword choices. I'm out of ideas and nothings working.
    >>
    No.13189 [Report]
    ▶
    File: ComfyUI_131011_.png
    (1.37 MB, 1024x1024) ▶
    1437670
    >>13179
    the tag to do it is: "hollow_head".

    but honestly it doe snot work very well. If you really like that things and have a bunch of refernce pics (even a couple) I can try making a specific LoRA for it.
    >>
    No.13190 [Report]
    ▶
    File: ComfyUI_93395_.png
    (1014 KB, 1024x1024) ▶
    1038662
    >>
    No.13191 [Report]
    ▶
    File: ComfyUI_130999_.png
    (1.37 MB, 1024x1024) ▶
    1438631
    >>13179
    now i can make it do half head like this.
    >>
    No.13192 [Report]
    ▶
    File: ComfyUI_131015_.png
    (1.12 MB, 1024x1024) ▶
    1177556
    >>13191
    But if I try to have a pic from above I get this.

    I did not have good reference pics to do it.
    >>
    No.13193 [Report]
    ▶
    File: ComfyUI_131046_.png
    (1.41 MB, 1024x1024) ▶
    1478983
    >>13192
    >>
    No.13194 [Report]
    ▶
    File: ComfyUI_131047_.png
    (1.39 MB, 1024x1024) ▶
    1458840
    >>13193
    >>
    No.13196 [Report]
    ▶
    File: ComfyUI_131131_.png
    (1.26 MB, 640x1536) ▶
    1324187
    I am making some models with the spit.

    First the model for playing with the spit.

    https://mega.nz/file/EBhAlISA#8iKUeqT1M8_XRpdtTlbdnBjGhC2780a9YAkfXUmqQZ8


    Strength:0.9
    CFG:4-5 (lots of artifacts at higher CFG)

    Trigger words: steel spit,

    Additional trigger words:

    arm support,

    titjob, grabbing own breast, between breasts,

    crotch rub, squatting, masturbation, female masturbation, mons pubis,
    >>
    No.13197 [Report]
    ▶
    File: ComfyUI_131171_.png
    (1.15 MB, 768x1344) ▶
    1208626
    >>13196

    Prompt:
    steel spit, 1girl, solo, long hair, breasts, looking at viewer, blush, smile, , large breasts, brown hair, navel, holding, , sitting, closed mouth, nipples, , nude, sweat, , , pussy, , collar, arm support, , kubo tite,
    kitchen background,

    masterpiece,best quality,very aesthetic,highres,absurdres,


    Negative prompt:
    3D, CGI, render, photo, text, watermark, low-quality, signature, moiré pattern, downsampling, aliasing, blurry, , blur, jpeg artifacts, compression artifacts, pixelated,
    censored, mosaic censoring, bar censor, blurry, bloom, deformed hands, head out of frame, watermark, worst quality, low quality, jpeg artifacts,
    multiple views, panels, speech bubble, text, bad hands, six fingers, deformed hands, english text,
    >>
    No.13198 [Report]
    ▶
    File: ComfyUI_131164_.png
    (1.17 MB, 768x1344) ▶
    1229369
    >>13197
    Prompt:
    steel spit, 1girl, solo, , breasts, large breasts, , , , nipples, , , pussy, , high heels, sandals, , squatting, grabbing own breast, ninja outfit, , shirow masamune,

    masterpiece,best quality,very aesthetic,highres,absurdres,




    Negative prompt:
    3D, CGI, render, photo, text, watermark, low-quality, signature, moiré pattern, downsampling, aliasing, blurry, , blur, jpeg artifacts, compression artifacts, pixelated,
    censored, mosaic censoring, bar censor, blurry, bloom, deformed hands, head out of frame, watermark, worst quality, low quality, jpeg artifacts,
    multiple views, panels, speech bubble, text, bad hands, six fingers, deformed hands, english text,
    >>
    No.13199 [Report]
    ▶
    File: ComfyUI_131185_.png
    (1.19 MB, 1024x1024) ▶
    1251955
    I forgot to mention. THis was trained for Illustrious.

    THe pics provided are made with WAI 14.0
    They are madde in comfyUI and you can just drag and drop them to get the workflow with all parameters I used.



    Prompt:
    steel spit, 1girl, solo, , breasts, skirt, large breasts, , , nipples, , , pussy, , , , squatting, , asanagi,
    school outfit, shirt, thighhighs, crotch rub, squatting, masturbation, female masturbation, mons pubis,

    school background,
    masterpiece,best quality,very aesthetic,highres,absurdres,



    Negative prompt:
    3D, CGI, render, photo, text, watermark, low-quality, signature, moiré pattern, downsampling, aliasing, blurry, , blur, jpeg artifacts, compression artifacts, pixelated,
    censored, mosaic censoring, bar censor, blurry, bloom, deformed hands, head out of frame, watermark, worst quality, low quality, jpeg artifacts,
    multiple views, panels, speech bubble, text, bad hands, six fingers, deformed hands, english text,
    >>
    No.13210 [Report]
    ▶
    File: ComfyUI_131329_.png
    (1.08 MB, 832x1216) ▶
    1130450
    New LoRA for Illustrious.
    Vertical impalement.

    https://mega.nz/file/kR4XlIIb#LeiW2DFsUC_DklHdjuId5RQYBDQ6EEEs4X5hUB7Hv14

    impaled_vertical-IL

    Strength:1-1.1
    CFG: any, I use 7

    Trigger words: impaled vertical, spit from mouth, steel spit,

    For headless: spit from neck, headless,

    additional: blood, defloration,
    >>
    No.13211 [Report]
    ▶
    File: ComfyUI_131255_.png
    (550 KB, 832x1216) ▶
    564085
    >>13210
    Promt:

    steel spit, spit from mouth, impaled vertical, 1girl,

    solo, long hair, breasts, skirt, large breasts, shirt, thighhighs, , nipples, , , pussy, , high heels, , , grabbing own breast, school uniform, ,
    school background, realistic,
    tongue, tongue out, ahegao, rolling eyes, school background, detailed background,

    masterpiece,best quality,very aesthetic,highres,absurdres,




    Negative prompt:
    3D, CGI, render, photo, text, watermark, low-quality, signature, moiré pattern, downsampling, aliasing, blurry, , blur, jpeg artifacts, compression artifacts, pixelated,
    censored, mosaic censoring, bar censor, blurry, bloom, deformed hands, head out of frame, watermark, worst quality, low quality, jpeg artifacts,
    multiple views, panels, speech bubble, text, bad hands, six fingers, deformed hands, english text,
    >>
    No.13212 [Report]
    ▶
    File: ComfyUI_131244_.png
    (965 KB, 832x1216) ▶
    988481
    >>13211
    Prompt:
    steel spit, spit from mouth, impaled vertical, 1girl,

    solo, long hair, breasts, skirt, large breasts, shirt, thighhighs, , nipples, , , pussy, , high heels, , , grabbing own breast, school uniform, asanagi,

    masterpiece,best quality,very aesthetic,highres,absurdres,




    Negative prompt:
    3D, CGI, render, photo, text, watermark, low-quality, signature, moiré pattern, downsampling, aliasing, blurry, , blur, jpeg artifacts, compression artifacts, pixelated,
    censored, mosaic censoring, bar censor, blurry, bloom, deformed hands, head out of frame, watermark, worst quality, low quality, jpeg artifacts,
    multiple views, panels, speech bubble, text, bad hands, six fingers, deformed hands, english text,
    >>
    No.13213 [Report]
    ▶
    File: ComfyUI_131357_.png
    (853 KB, 832x1216) ▶
    874027
    Prompt:
    steel spit, spit from neck, headless, impaled vertical, 1girl, solo, breasts, large breasts, nipples, nude, shiny, arms behind back, bound, bdsm, bondage, rope, frogtie, , from side,
    3d,

    masterpiece,best quality,very aesthetic,highres,absurdres,



    Negative prompt:
    3D, CGI, render, photo, text, watermark, low-quality, signature, moiré pattern, downsampling, aliasing, blurry, , blur, jpeg artifacts, compression artifacts, pixelated,
    censored, mosaic censoring, bar censor, blurry, bloom, deformed hands, head out of frame, watermark, worst quality, low quality, jpeg artifacts,
    multiple views, panels, speech bubble, text, bad hands, six fingers, deformed hands, english text,
    >>
    No.13214 [Report]
    ▶
    File: ComfyUI_130817_.png
    (1.09 MB, 1024x1024) ▶
    1147989
    New LoRA for Illustrious.
    Horizontal impalement. Also includes spit roast.

    https://mega.nz/file/kFp13apL#91I3ETvFD65JRdz6_sesGb9KVsb3IzgmqDlgJb5zHlo

    roasted-IL_horizontal
    Strength:1
    CFG: whatver, I use 7

    Trigger words: spit from mouth, impaled horizontal,steel spit,
    Additional roasting tags: roasted over fire, BBQ head BBQ body, steaming body,
    >>
    No.13215 [Report]
    ▶
    File: ComfyUI_130764_.png
    (925 KB, 1344x768) ▶
    947746
    >>13214
    Prompt:
    hyuuga_hinata, naruto_(series), 1girl, black_hair, purple_eyes,

    high heels,
    spit from mouth, impaled horizontal, steel spit, , 1girl, solo, breasts, , open mouth, large breasts, nipples, , , tongue, tongue out, , , , , arms behind back, bound, bdsm, bondage, , , cuffs, ,

    masterpiece,best quality,very aesthetic,highres,absurdres,




    Negative prompt:
    3D, CGI, render, photo, text, watermark, low-quality, signature, moiré pattern, downsampling, aliasing, blurry, , blur, jpeg artifacts, compression artifacts, pixelated,
    censored, mosaic censoring, bar censor, blurry, bloom, deformed hands, head out of frame, watermark, worst quality, low quality, jpeg artifacts,
    multiple views, panels, speech bubble, text, bad hands, six fingers, deformed hands, english text,
    >>
    No.13216 [Report]
    ▶
    File: ComfyUI_130869_.png
    (1.1 MB, 1344x768) ▶
    1155056
    Prompt:
    , maid,
    , cannibalism, steaming body, BBQ head, BBQ body, roasted over fire,
    spit from mouth, impaled horizontal, steel spit, , 1girl, solo, breasts, , open mouth, large breasts, nipples, , , tongue, tongue out, , , , , arms behind back, bound, bdsm, bondage, , , cuffs, ,

    masterpiece,best quality,very aesthetic,highres,absurdres,




    Negative prompt:
    3D, CGI, render, photo, text, watermark, low-quality, signature, moiré pattern, downsampling, aliasing, blurry, , blur, jpeg artifacts, compression artifacts, pixelated,
    censored, mosaic censoring, bar censor, blurry, bloom, deformed hands, head out of frame, watermark, worst quality, low quality, jpeg artifacts,
    multiple views, panels, speech bubble, text, bad hands, six fingers, deformed hands, english text,
    >>
    No.13217 [Report]
    ▶
    File: ComfyUI_130850_.png
    (1.41 MB, 1344x768) ▶
    1482801
    >>13216
    Prompt:
    shirow masamune, maid, from below, from below,
    , cannibalism, steamiing body,
    spit from mouth, impaled horizontal, steel spit, , 1girl, solo, breasts, , open mouth, large breasts, nipples, , , tongue, tongue out, , , , , arms behind back, bound, bdsm, bondage, , , cuffs, ,

    masterpiece,best quality,very aesthetic,highres,absurdres,







    Negative prompt:
    3D, CGI, render, photo, text, watermark, low-quality, signature, moiré pattern, downsampling, aliasing, blurry, , blur, jpeg artifacts, compression artifacts, pixelated,
    censored, mosaic censoring, bar censor, blurry, bloom, deformed hands, head out of frame, watermark, worst quality, low quality, jpeg artifacts,
    multiple views, panels, speech bubble, text, bad hands, six fingers, deformed hands, english text,
    >>
    No.13218 [Report]
    ▶
    File: ComfyUI_130829_.png
    (1020 KB, 1344x768) ▶
    1045108
    >>13217
    Prompt:
    ayaswan, maid, from below, from below,

    , BBQ body, BBQ head, cannibalism, steamiing body,
    spit from mouth, impaled horizontal, steel spit, , 1girl, solo, breasts, , open mouth, large breasts, nipples, , , tongue, tongue out, , , , , arms behind back, bound, bdsm, bondage, , , cuffs, ,

    masterpiece,best quality,very aesthetic,highres,absurdres,





    Negative prompt:
    3D, CGI, render, photo, text, watermark, low-quality, signature, moiré pattern, downsampling, aliasing, blurry, , blur, jpeg artifacts, compression artifacts, pixelated,
    censored, mosaic censoring, bar censor, blurry, bloom, deformed hands, head out of frame, watermark, worst quality, low quality, jpeg artifacts,
    multiple views, panels, speech bubble, text, bad hands, six fingers, deformed hands, english text,
    >>
    No.13241 [Report]
    ▶
    File: ComfyUI_02803_.png
    (1.05 MB, 1344x768) ▶
    1097459
    >>13218 Hero!!

    Of course I had to try this on my favourite spy.
    I do seem to have a bit of a problem of cooking her head though.
    >>
    No.13249 [Report]
    ▶
    >>13241
    it does. I believe this is due character being a certain color is more embedded part thant their body being a ccertain color.

    FOr this reason many concepts like taking a body to pieces does not work well.


    If there is enough demand (aka someone gives me a few USD on civitai) i can train aversion composed entirely of girls with cooked faces. That should be more functional.
    >>
    No.13280 [Report]
    ▶
    made a short story

    https://files.catbox.moe/iz9b5w.zip
    >>
    No.13296 [Report]
    ▶
    >>13280
    Would love to see it, but catbox seems to be giving a lot of problems lately
    >>
    No.13305 [Report]
    ▶
    File: Nun_3.png
    (1.2 MB, 832x1216) ▶
    1263261
    >>13296
    made a reupload
    https://www.mediafire.com/file/gvg02h2q67pge4f/nun.zip/file


    hope this works for you
    >>
    No.13309 [Report]
    ▶
    >>13215
    Hi I tried this, and cooked/uncooked head is still difficult to seperate.maybe when you train is lora, annd uncooked head prompt to differentiate two type of heads?
    >>
    No.13310 [Report]
    ▶
    File: 00003-1530145052.png
    (2.26 MB, 1792x1224) ▶
    2374985
    fantastic
    >>
    No.13312 [Report]
    ▶
    >>13305
    Thanks! And they look awesome! Very consistent as well!
    >>
    No.13330 [Report]
    ▶
    >>13310
    Thanks! I have seen your gens on pixiv. Nice job!

    >>13312
    TY, you remined me I need to retouch it in photoshop for more consistency

    >>13309
    yeah, the new version had synthithic cooked head added ot the dataset.
    >>
    No.13427 [Report]
    ▶
    >>13418
    I kinda love this artstyle. Could you post some more pics with it?
    Please :D
    >>
    No.13443 [Report]
    ▶
    >>13418
    >>13440
    Kurzgesagt style for guro?

    Lol. Diagrams that explain how to do horrbile guro things.
    >>
    No.13449 [Report]
    ▶
    >>13443
    I definitely wouldn't mind.
    >>
    No.13548 [Report]
    ▶
    >>13139
    This is so insanely good. Please share your prompt, settings and other stuff!
    I need to see more like this!
    >>
    No.13864 [Report]
    ▶
    File: Lina.jpg
    (159 KB, 1568x2016) ▶
    163547
    impaled vertical lora is awesome
    >>
    No.13912 [Report]
    ▶
    >>13864
    that looks nice.

    >>
    No.13913 [Report]
    ▶
    new model out in the new thread

    >>13910
    >>
    No.22837 [Report]
    ▶
    can you reshare the g model v1.0? i find it better in some aspects than v2.0
    >>
    No.23137 [Report]
    ▶
    Amazing work! Do you take requests?
    [Top]
    Delete post: []
    First
    [1]
    		Last