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.
  • 114 unique users in the last 30 minutes (including lurkers)
    First
    [1]
    		Last
    File: 00054-535290644.png
    (2.37 MB, 1200x1600) ▶
    2487129
    No.12868 [Report] AS
    ▶
    illustrious
    ttps://files.catbox.moe/qm8hhq.Zip
    pony
    ttps://files.catbox.moe/bzrnve.Zip
    dataset for anyone interested
    ttps://files.catbox.moe/fn6heo.Zip
    password : transverse
    catbox would not work in some region. You can use VPN.

    After making sagittal plane cutting v2 using multi-phase training, block locking, reg images, I thought I could make transverse plane cutting v2.1 using the same technique. But as it turned out the training data was not good as I thought it was so I had to fix it, so it ended up becoming v3.
    For next I'm considering various impalement LoRA but I'm not sure whether it will ever happen.
    Some sample images starting with Frenda who just can't catch a break.

    pony
    >>
    No.12869 [Report]
    ▶
    File: 00080-1516404582.png
    (2.61 MB, 1600x1200) ▶
    2735405
    illustrious
    >>
    No.12870 [Report]
    ▶
    File: 00042-1771773157.png
    (2.28 MB, 1600x1200) ▶
    2386801
    getting fucked after being chopped into two
    pony
    >>
    No.12871 [Report]
    ▶
    File: 00094-2374398015.png
    (2.56 MB, 1200x1600) ▶
    2689557
    pony
    different aspect ratio this time
    >>
    No.12872 [Report]
    ▶
    File: 00100-160686824.png
    (2.32 MB, 1600x1200) ▶
    2428236
    illustrious
    >>
    No.12873 [Report]
    ▶
    File: 00116-2270931680.png
    (2.78 MB, 1600x1200) ▶
    2918160
    illustrious
    tried some realistic checkpoints. doesn't work all that well. Need really high LoRA strength.
    >>
    No.12874 [Report]
    ▶
    File: 00097-522777711.png
    (2.75 MB, 1600x1200) ▶
    2879191
    by adding 'lower torso' in positive prompt and 'upper torso' in negative prompt, you can generate an image with only the lower part.(gacha needed)
    pony
    >>
    No.12875 [Report]
    ▶
    File: 00045-771720946.png
    (2.29 MB, 1600x1200) ▶
    2404311
    illustrious
    it's really hard to get the spine right, so if you don't want some heavy gacha, just leave it out.
    >>
    No.12876 [Report]
    ▶
    File: 00044-3169089473.png
    (1.89 MB, 1600x1200) ▶
    1984433
    lower torso getting fucked
    pony
    >>
    No.12877 [Report]
    ▶
    File: 00068-2013649492.png
    (2.1 MB, 1600x1200) ▶
    2204337
    illustrious
    >>
    No.12878 [Report]
    ▶
    File: 00052-2218248429.png
    (2.47 MB, 1200x1600) ▶
    2587458
    on the opposite, add 'upper torso' in positive prompt and 'lower torso' in negative prompt to get only upper body.(gacha needed)
    some unfortunate magical girl
    pony
    >>
    No.12879 [Report]
    ▶
    File: 00082-1590838743.png
    (2.41 MB, 1200x1600) ▶
    2522078
    illustrious
    >>
    No.12880 [Report]
    ▶
    File: ComfyUI_00060_.png
    (2.65 MB, 1024x2560) ▶
    2773718
    tried somewhat complicated stuff using comfyUI to show where the rest of her body go
    only illustrious this time
    >>
    No.12881 [Report]
    ▶
    File: 00016-1328270513.png
    (2.92 MB, 1600x1200) ▶
    3060340
    some samurai getting chopped
    pony
    ok pony6 sucks, i won't make any more LoRA for pony6
    >>
    No.12882 [Report]
    ▶
    File: 00078-4027140561.png
    (2.63 MB, 1600x1200) ▶
    2762690
    pony
    tried with some realistic checkpoint
    >>
    No.12883 [Report]
    ▶
    File: 00100-538072901.png
    (2.64 MB, 1600x1200) ▶
    2772129
    illustrious
    >>
    No.12884 [Report]
    ▶
    File: 00055-538072856.png
    (2.7 MB, 1600x1200) ▶
    2834596
    another illustrious
    always better result with less gacha
    >>
    No.12885 [Report]
    ▶
    File: 00057-3555556043.png
    (2.76 MB, 1600x1200) ▶
    2889957
    now for some western knights
    pony
    >>
    No.12886 [Report]
    ▶
    File: 00078-3555556043.png
    (2.61 MB, 1600x1200) ▶
    2733087
    pony
    attempt with some realistic checkpoint
    >>
    No.12887 [Report]
    ▶
    File: 00076-3958357455.png
    (2.94 MB, 1600x1200) ▶
    3086332
    illustrious
    >>
    No.12888 [Report]
    ▶
    File: 00058-3958357437.png
    (2.67 MB, 1600x1200) ▶
    2798480
    illustrious
    >>
    No.12889 [Report]
    ▶
    File: 00080-3246167623.png
    (2.69 MB, 1200x1600) ▶
    2824283
    here are some tests combining transverse plane cutting with sagittal plane cutting. It was pretty impossible with pony, and difficult even with illustrious.
    saint armor
    illustrious
    >>
    No.12890 [Report]
    ▶
    File: 00031-3337324351.png
    (2.64 MB, 1200x1600) ▶
    2771649
    knight
    illustrious
    >>
    No.12891 [Report]
    ▶
    File: 00044-1285455465.png
    (2.46 MB, 1200x1600) ▶
    2579281
    harpy
    illustrous
    >>
    No.12892 [Report]
    ▶
    File: ComfyUI_00023_.png
    (5.46 MB, 2616x1472) ▶
    5721525
    used comfyUI for more complicated composition.
    Knights killing each other.
    It turns out making two characters face each other was ten times more difficult than getting the guro right.
    illustrious
    >>
    No.12893 [Report]
    ▶
    File: ComfyUI_00115_.png
    (5.47 MB, 2616x1472) ▶
    5734689
    illustrious
    >>
    No.12895 [Report]
    ▶
    File: ComfyUI_00017_.png
    (5.17 MB, 2616x1472) ▶
    5423894
    naked amazon version
    illustrious
    >>
    No.12908 [Report]
    ▶
    the fact people still use and train for pony is stupid and insane. illustrious and noobai just shit all over it in terms of quality and output.

    would you be cool with posting the dataset? i'd be willing to train it on chroma (i trained the sagittal lora on chroma in the other thread)
    >>
    No.12909 [Report]
    ▶
    >>12908
    sorry, missed that you posted the dataset too. i'll play around with it and see what i can do
    >>
    No.12974 [Report]
    ▶
    >>12868

    Nice! Thanks for sharing! :)

    >For next I'm considering various impalement LoRA but I'm not sure whether it will ever happen.

    I've still got the datasets that I've prepared in collaboration with Entropy. Sadly he has gone dark, so the Lora was never finalised, but I think we were very close back then, at least for vertical impalements, but I had also collected stuff for horizontal impalement variants. If you're interested in the datasets please feel free to hit me up on pixiv (user id: 15925390) :)
    >>
    No.12983 [Report]
    ▶
    File: ComfyUI_temp_ljdye_00018_.png
    (4.06 MB, 1536x1536) ▶
    4257313
    >>12868
    Really amazing model, great job! The output is very clean, and tags are clear and responsive. This was done with just t2i, no other lora, only some quality embeddings.
    [Top]
    Delete post: []
    First
    [1]
    		Last