This model aims to help generate a variant of bursting breasts: breasts that are about to completely burst of clothes and have been pressed into an unusual shape because of that. Basically, the before image of exploding clothes.
I intend to improve and update this with more views, clothing, etc. as I learn more about training models.
V2.2
Another small expansion of the dataset and some retraining. The model is a little more responsive to tag weights and can sometimes do a clothed concept of itself (i.e. breasts bursting out of a bra while wearing a collared shirt over).
My image saving node messed up, so the example images may not be available for a while. The metadata didn't save to the images themselves and I may or may not get around to manually inputting the generation info for each.
V2.1
Slightly expanded the dataset and made an Illustrious model.
V2.0
Retagged the dataset with a newer tagging model and slightly modified how the tags work to help with the way breasts spill out. Most important is replacing atb with bursting rbeasts since that produced better results in all my trials. In short, cleavage is breasts spilling from the top, sideboob is from the side, and underboob is from below. In many cases, it's just as important to specify the locations where you don't want spillage as it is to the locations where you do. Tags also seem more stable, so increasing the weight of them even up to 2+ has produced good results.
XL V1.4
Finally got around to making this for SDXL. I'd recommend running the SDXL LoRA at 1 normally. SDXL does a way better job of following prompts, so a higher weight seems to add in a style more than it improves the image. Some tags also introduce a style. I'll be trimming the dataset and re-tagging it shortly to try and eliminated this.
V1.4
Expanded the dataset by 80 or so images. I also re-tagged every image and combed through the tags to make certain everything looked right. Poses, sizes, and prompts are more consistent as a result. Now combinations like cleavage in the prompt and underboob in the negative should help produce results with only cleavage spilling out of clothes. Likewise, including both cleavage and underboob eliminates most sideboob instances. I also nudge the upper limit of the enormous breasts tag and the wide hips and thick thighs tags are slightly larger on average.
I'd recommend running this LoRA at 1.3 normally and around 1.3-1.5 when other LoRAs are involved. This LoRA is heavily impacted by the canvas size and orientation. Wider images allow for larger sizes but less diversity, and a more narrow canvas produces more dynamic poses but may force larger tags to generate smaller breast sizes. Don't be afraid to try all sorts of canvas layouts.
The posted images were upscaled with StableDifusion's built in Anime upscaler to x2 and ran through GFPGAN at a visibility of 0.6.
V1.3
Greatly expanded the dataset to help with tag recognition. I also changed some tags around to make the enormous breasts and gigantic breasts tags more more consistent.
V1.2
The from side tag now consistently generates images and the from behind tag will generate most of the time. I recommend running this at 1.3 weight. The biggest change is the addition of the enormous breasts tag. This was included to allow for larger sizes than gigantic without making the gigantic breasts tag inconsistent. This new tag favors wider images, so if it seems to generate a smaller size than anticipated, try a 640x512 or 768x512 ratio. If you can't quite get a pose right with view tags, try some helper tags. For instance, I had to use looking away and backboob to get a good from behind pose.
V1.1
I made the from-side tag more consistent and finally got the from-behind tag to generate something coherent. From-behind is mostly a proof-of-concept right now. Also made the breast size distinction more impactful.
V1.0
Further refined the dataset and added a few more images. Also refined the tagging so side view images could be recognized. The image generation for all sizes from the front seems to be consistent now, but images from the side are hit or miss. The from-behind view is still unsupported.
V0.2
Refined the dataset so generation is more consistent and breast sized are more defined.
To do:
Fix trigger wordFix poor hand generationFix poor nipple generationFix multiple breast generationAdd side view supportAdd rear view support
Add top/bottom view supportBe able to specify if breasts burst from top or bottomTry and support smaller breast sizesRemove SDXL LoRA adding a style
Support torn clothes better
Description
FAQ
Comments (18)
NOW A PONY MODEL?!?!? Bro Slow your cooking down XD, it's too good Lol
My LoRA production ability has been unbound now that I have time for it. I can cook to my heart's content
@reggiemccorthy1457 XD good for you bud, love these new sdxl one's.
atb is not a valid trigger. Correct bursting breasts . In addition, for stability, I advise prompts undersized clothes, bra,clothes press breasts, for effectiveness you can increase either the weight of lora to 1.5 or the value of the hints (cleavage), cleavage cutout, ((undersized clothes)), bra,((clothes press breasts))
How is atb not a valid trigger? I am under the impression that you can train a LoRA to use any valid characters as a trigger. If that's not correct, could you inform as to how to correct it?
@reggiemccorthy1457 atb is a trigger, it works for me in js2prony
@reggiemccorthy1457 Here's for comparison https://civitai.com/images/15834197 It can be seen that with a weight of lora 1.4 atb does not manifest itself in any way, but bursting breasts exhibit the property of lora to a significant extent.
Here is lora with and without atb https://civitai.com/images/15827792, https://civitai.com/images/15827569
Here is lora with and without bursting breasts https://civitai.com/images/15827616, https://civitai.com/images/15827821
@pokezilla39966 Please show me some art where this lora with a high weight will be separately with an atb trigger and a bursting breasts trigger.
@fansay ok, but i have to say, using it at a lora weight of 1.4 absolutely knocks down the quality, so i'll use 1.0 lora weight, give me 10 minutes to an hour or 2 for further testing
@fansay I tried using the image I always use to test this model. In order, I tried: no triggers, atb only, bursting breasts only, atb and bursting breasts. Image post by reggiemccorthy1457 | Civitai
I got everything to have an effect, including atb. Though bursting breasts definitely has more of a pull than I'd like. I also tried copying your generation parameters for the images in the first link and got entirely different results. I really have no idea why atb doesn't work for you.
@reggiemccorthy1457 it works for me, https://civitai.com/posts/3535533 i used with atb and with bursting breasts, admittedly, bursting breasts also has stronger effect, oddly enough than ATB.
@reggiemccorthy1457 The simple common promts 1 girl, upper body, bra First adding atb and then replacing it with bursting breasts. And show the art. No huge breasts, because this will already affect the result. Or in Script select X/Y/Z plot ->Prompt S/R ->in X values atb,bursting breasts
At the same time, the promts should be this: 1 girl, upper body, bra, atb
@pokezilla39966 I will repeat for you. The simple common promts 1 girl, upper body, bra First adding atb and then replacing it with bursting breasts. And show the art. No huge breasts, because this will already affect the result. Or in Script select X/Y/Z plot ->Prompt S/R ->in X values atb,bursting breasts
At the same time, the promts should be this: 1 girl, upper body, bra, atb
@fansay this may work for you bud, but it may work differently for others depending solely on what checkpoint and service they use, for example: ComfyUi, Auto1111 or any other Ai generation service. for me it's Auto1111 with js2prony, hi-res fix: using 4x-animeSharp, no image upscale, 20 hi-res steps, 0.4 denoising. if it works for me with how i use my setup, then i don't see why i should change my generation lines to use your recommended. (sorry, you kinda rubbed me the wrong way with how you are saying how to use the generation prompts). i will admit the script part helped, always wondered where that was lol XD
@pokezilla39966 Maybe, maybe not. These are all words. Show clearly with one trigger and with another trigger with ordinary promts, as an example that I offered.
@fansay i getcha, but your script part was more helpful, i never knew how to work script and now i do, and it's a gamechanger for previous idea's i had months prior, so in a way, you did help me, thanks :D
the SDXL version works well on anime models, but doesn't have much or any effect at all on photorealistic models such as jibMixRealisticXL, JuggernautXL, epicRealismXL_7. Hope you can give it a try on these realistic models and figure out a fix for it.
I wouldn't be surprised if it's just the dataset. All the images this was trained on were anime-like illustrations. I'd say that combined with the base model for training being an anime SDXL model as well keeps it from effectively working on realistic models.
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.