Update: I've trained a new version with a much better training setup and I also trimmed the dataset and cut it into 3 categories to give much more control over the size and type of bulge, plus one version that includes all 3 sets which I highly recommend mixing in with other versions to improve consistency.
Version Nice: dataset trimmed for images with nicely detailed bulges that aren't ridiculously huge (still pretty big though, I might make another version for better smaller bulges)
Version Absurd: selection of some of the most absurd insertions/penetrations. This one especially benefits from mixing version "All" with it, something like absurd:0.8 all:0.6 will give very consistent outputs. I tried merging them into one lora but every method I tried to do so gave vastly inferior results compared to putting both loras in prompt.
Version Pectoral: images with bulges reaching up to chest. Intended to be for impossible penetrations going right through the chest instead of the stomach bulge extending up to chest level, but it's ambiguous.
Tips: Don't forget to add vaginal/anal, it'll vastly improve the penetration part. Also, I recommend adding fat or pregnant in negative to improve the shape of the bulge.
LORA trained on a large set of images focused on large penetration, more specifically large stomach distention. I tried going for versatility and minimal visual impact. High epochs with high weight can get very weird, I'd recommend starting with epoch 4 and 0.8 or lower weight (v7 seems fine with higher). Even then this is meant for extreme content, if you just want normal large penetration many models can do that very well already.
For more absurdity, try things like absurd stomach bulge, all the way up torso, or ask for sideways viewpoint. I've added a lot of tags to training data, but it doesn't particularly react more strongly to exact tags, just try things. Seems pretty difficult to train a LORA that can react to many specific things like the shape of the stomach bulge, I tried but without much success, at most minimal changes on occasion when forcing it with 1.5+ emphasis. (things like ribbed or spiky dildos, or cone-shaped object insertion)
Combines well with other LORAs, as demonstrated in the last example image. Extreme insertion can improve the penetration part, tentacle bulge gives interesting shape to the bulge, and autofellatio (and maybe other futa ones) helps the AI draw absurd bulges by thinking it might be a dick at first.
Description
I've noticed v6 changing the artstyle on some (usually smaller) prompts so I lowered the learning rate, seems better overall.
FAQ
Comments (35)
where can i find that gape model you are using in the promos
It's a mix of well known models, regular abyssOrangeMix2 or grapefruit will get you similar results. If you want to try replicating this exact merge, the mix is:
(BloodOrangeMix + Basil Mix) + (7th anime 1.1 + Abyss2hard) and then 30% gape on top. Gape merge at the end is difference between gape60 and NAI, every other merge is block merge (using Merge Block Weighted extension) with the top preset option (grad_v), except Blood+Basil is "1,0.9,0.7,0.5,0.3,0.1,0,0,0,0,0,0,0,0,0,0,0,0,0,0.1,0.3,0.5,0.7,0.9,1". base_alpha is 0 on every merge.
Having issues getting that merge right. Any chance you can upload the model? Awesome work by the way :)
@Zygany Eh, I don't know, it's just one of many random merges I made. I don't really like the flooding of everyone's custom mixtapes on the site either.
I made a mistake in that recipe, the gape diff is with NAI, not SD 1.5. Also the block merge for "bloodbasil" might be slightly different, specifically "1,0.9,0.7,0.5,0.3,0.1,0,0,0,0,0,0,0,0,0,0,0,0,0,0.1,0.3,0.5,0.7,0.9,1"
@wewewew the number is hard to understand, what is it buddy?
@endlessvn2604 It's used in the "Merge Block Weighted" extension.
Excellent stuff! Thanks for the work!
Will do some testings and post some results here :)
any chance you can upload the merged gape model here?
anyhow thanks for the great effort!
Could you train one on Sparrow's art next? :>
There are a few of them in the training set, but most of Sparrow's art is way too incomprehensible for the AI to learn.
Hi there, can I ask where to get lora:Jordan_3, aki_v4 and what is 002 Models? Or just give me the full name, thank you!
Jordan_3 (background focused lora that also seems to increase general coherency) appears to have been deleted, I've reuploaded it here https://files.catbox.moe/ktii8u.safetensors
aki lora (aki99) you can find here https://gitgud.io/gayshit/makesomefuckingporn
@wewewew awesome! Thank you so much!
@wewewew Jordan_3 link lost ，sad
@levi_rem Works fine, maybe the site was down for a bit?
where can i find that 002+gape30 model
I posted the recipe in another comment, but here's a copy paste.
(BloodOrangeMix + Basil Mix) + (7th anime 1.1 + Abyss2hard) and then 30% gape on top. Gape merge at the end is difference between gape60 and NAI, every other merge is block merge (using Merge Block Weighted extension) with the top preset option (grad_v), except Blood+Basil is "1,0.9,0.7,0.5,0.3,0.1,0,0,0,0,0,0,0,0,0,0,0,0,0,0.1,0.3,0.5,0.7,0.9,1". base_alpha is 0 on every merge.
I verified the recipe just now and updated the correct order of models to merge (caus it does matter which model you put in which slot in the MBW extension)
Again though, the reason I'm not posting the model is because it's some random merge that I didn't do much testing for or anything, you'd probably get better results trying different merge methods or just using something else.
@endlessvn2604 It's a difference merge between gape60 and NAI, difference merge is available in the regular checkpoint merge tab, you put the merge you've done in slot A, gape60 in slot B, and NAI in slot C. Set it to 30 and it will add 30% of the difference (B-C) into A.
@wewewew thank you, that was an ultimate model!
@wewewew why dont you submit your model in demo to civitai,I tried some other model but no one goods as demo.your demo is amazing~~thanks~
I don't have enough RAM to do this...
Yeah can @wewewew can you post the model?
@wewewew can you post your model?
could you please make a fanny-packing model
can someone send me to the jordan lora?
Can it do unbirth? If not that would be my dream model.
What model was used on the main images? Where to find?
perfect！I really like this mod, may I ask your first picture's tags and other information?
You can see image's prompt and configs by clicking on it.
@wewewew got it! Thank you
The stomach bulge cant covered by anything,stomach must be exposed when I use the lora，it still a perfect model
Do you have suggestions how to make this work with models that's not based on NAI or gape60? I have been trying to create hyper penetration LoRA for realistic models but I think LoRA isn't enough to do that job because I have to merge gape60 or similar lora to make it work.
I've not really tried using realistic models, although semi-realistic 3D stuff works alright. You could try making a diff lora of gape, using the supermerger extension, extract the difference between gape60 and NAI (leave both ratios at 1) to a lora with not too high dim, like 32 or so, it usually integrates much better with various models than merging the models themselves.
Two months ago, I tried that but with 256 and 512 dim but I will try smaller dim. I also could try to train that LoRA with low LR. Thanks!
would love to see that!
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.