A model for large breasted waifus or semi-realistic characters.
V6 Changelog 2023/06/03:
Considering this was my first and most popular LoRA, I figured it was time to improve on it. Probably the last version too. Hyperfusion is more or less the successor to this.
Less harmful to the overall style, but slightly smaller sizes overall, feels like a good balance
Wider variety of breast shapes in the dataset
Some breast shape tags were added, but hard to say how helpful they are
Uses LoRA LoCon, but does not require any additional extensions to run
Changed "large breasts" to "big breasts" to match my hyperfusion model tags
The model prefers resolutions around 640, but can do up to 768 to a degree
Training tags attached under model download section
V5 Changelog:
v5 now uses LoRA!
What changed since v4? I trained a breast size classifier to auto label the dataset which resulted in better control over breast size, and made it easier to generate larger sizes.
Note that the trigger words in v5 are ordered in the following from smallest size to largest:
breasts, large breasts, huge breasts, gigantic breasts
V4 Info:
the keyword for v4 remains hyperbreasts, if you want better control over the size use v5
Notes:
I used this to train my image tagging classifiers for breast sizes
https://github.com/huggingface/transformers/tree/main/examples/pytorch/image-classification
Description
Added some more variety in breast shapes to the dataset
Less harmful to the overall style than v5
Training Details:
2k images
16 epocs
learning rate 2e-4
text encoder LR 5e-5
base model Av3
clip skip 2
random flip
tag drop chance 0.1
bucketing at 768
dim 16
alpha 8
150 tokens
cosine with restarts
training with tags, tags attached next to model download
Kohya LoRA LoCon, does not require any additional extensions to use
FAQ
Comments (43)
How you trained a "breast size classifier"? Can you link a Youtube video or an article that explains similar process?
I used this, it's pretty straight forward, and similar to SD training in some ways
https://github.com/huggingface/transformers/tree/main/examples/pytorch/image-classification
I don't know of any guides, but you can try looking around.
Thanks dude!
v6 is infinitely better than v5, might actually be able to use it now
Nice!
such a funny lora
do you have any prompts that work with this for consistent results or its it just the ones featured in the images above?
Consistency is always hard, but if you are aiming for a specific body shape, I would refer to r-34. Find a similar image there, and try using the same tags. That's the best way to explore.
Bro I've tried all your models and they LITERALLY don't do anything, ever
Double check you are loading the LoRA correctly. Everyone else seems to be working fine.
I have every version of every LoRa installed, with the exception of Lacto... Every one of them works perfectly for me, generating images frequently... 🤔
More than happy to offer guidance, if you give me info on your software/hardware/prompting..?
was there not a bug with the latest stable diffusion update that made loras not work?
@midobatsu Good point, some LoRA's were broken in 1.0 an 1.1 of A1111. Before and after that is fine.
For anyone else having this problem, put it in your LYCORIS folder, then it started working for me.
@goblinmarket The latest A1111 (1.5+) unifies all LoRA, so upgrading should work too
Any tips for getting less of a large gap between the breasts? They looking good but I'm getting a lot of sternum lol
Does increasing the weight of `cleavage` help? Otherwise `breasts apart` in the negative with high enough weight might make a difference too, but I'm only making an educated guess.
So, as best I can tell the ai system I am using is reading the trigger words as separate tokens, thus not triggering the LoRA. I’m not really sure how to fix this.
Yea, the trigger words should be broken up by the tokenizer into separate tokens. Try one of the prompts in the images provided, if you still don't see it working it's likely the LoRA is not loading correctly.
@throwawayjm Yeah, it’s still not working for me. I’m not sure but it could be because I’m using the drawthings app to run my prompts, but I have no idea. I wouldn’t know how to fix the load, so this is probably beyond me for now. Still, thanks for the response. It looks great, hopefully I can figure out how to get it working in the future.
Hyper pussies please
nice one
Vital resource!
Any idea how to fix the weird nipples on this LORA? I'm getting a lot of inverted nipples and some that look downright infected! Other than that, it's amazing (with some trial and error)
Nipples have been hard to correct. Kinda like the hands problem. Using this LoRA below :1> strength probably makes the issue worse too. I guess you just have to get lucky, im not sure how to solve it.
@throwawayjm There is a lora called real nipples, if you use it and also read this guide https://www.reddit.com/r/StableDiffusion/comments/11u2p0u/lazy_guide_to_photorealistic_images/ you can get them to look pretty good! I use it like this: <lora:REAL-NIPPLES:0.3> perfect nipples
so i use sd 1.4, the recommended ckpt. and the generation data provided but im not getting near the to the result of any of those pictures. what besides that needs to be done to get to the examples? im new to sd so theres surely something im missing but i have no clue what it is :(
You just need this text somehwere in your prompt <lora:hyperbreasts_v6:1>, and make sure the LoRA file is in "/models/Lora" if you are using Automatic1111.
can i give an advice? If you want to use this LoRA for fully clothed characters, i recommend to combine this LoRA with rulles's Cutesy style for softer faces, and it works perfect, adding "<lora:Cutesy4-10:1> <lora:hyperbreasts_v6:1.3>" in the prompt, and "(off breasts:2), (cleavage:2)" in the negative prompt.
It has been months since you released v6 and I still can not get it to do anything I have even found and fixed some problems making all other loras work better but this one and only this one refuses to work on v6. I am going to redownload it again and give it one last try but man this will be like my 3rd or 4th time.
Usually the fix is upgrading Automatic1111 to a version above 1.2. Some older versions failed to load certain types of LoRAs.
@throwawayjm thankfully it did manage to work this time I can finally start using my favorite lora again
I can't seem to download this at all. It's throwing up a weird error but I think that's on civitai's side and not yours.
any chance this comes over to sdxl?
I tried with a few different datasets now. It's just too difficult to bother training right now. I've been waiting for someone to make a successful large dataset non-realistic finetune before venturing into SDXL as proof that it works at scale. It's probably possible to do, but I can make nearly the same finetune at 1024px on sd1.5 so why bother when SDXL is only marginally better.
@throwawayjm We know atleast that it is possible. NovelAI v3 is based off sdxl but thats about all we know
@loktar Yea I saw some examples from their new model. Who knows what tricks they used to train it =P
@throwawayjm Probably an H100 supercluster costing hundreds of thousands or millions to rent
Holy crap, despite the memes about larger-than-life breasts, this lora is unironically the best breast enhancing mod on civit that doesn't strip the character/person's clothes off when making them larger. When used at a lower percent, it just makes them bigger, it's amazing!
can you make the breast size gigantic, please,
Hyperfusion should be able to get bigger sizes. This model is older and less diverse.
this one you made is pretty aight, 8/10
Can we get this for QWEN?
Details
Files
Available On (2 platforms)
Same model published on other platforms. May have additional downloads or version variants.