Canceled by Civitai new ToS [2025]
When new Version is on i write it here - I dont Upload any RL LoRA here on Civitai more!
Free Version 1.2
1.2 Changelog:
Fixed broken face for many Checkpoints
Better stable generation for AnimateDiff
Enchant Output
All Preview has no Upscaled or ADetailer on!
LoRA Weight: 0.6-0.8
Sample Promts: (Promt List see below)
medium breast, pink nipples, normal_nipples, normal areola
big breast, pink nipples, saggy breast, big areola
Try many Times to Gen your Breasts.
Nipples and Forms are in eary stage!
Triggers:
Breast Size (Cup trigger - WiP)
a-cup breast
b-cup breast
c-cup breast
d-cup breast
Breast Size (Size trigger)
flat breast
small breast
medium breast
big breast
fake breast
bimbo breast
Breast form
torpedo_breast
saggy breast
Nipple Color
pink nipples
brown nipples
Nipple Style
normal_nipples
big nipples
inverted nipples
puffy nipples
Areola Size
small areola
normal areola
big areola
!OLD! Different Breast Types (triggerwords) (Beta 1&2)
flat_tits
small_tits
medium_tits
big_tits
fake_tits
bimbo_tits
Beta 2 Recommended strenght; 0.5-0.6!
Beta 1 Recommended strength: 0.6-0.8!
small boobs note: small boobs pictures were trained by shemales and femboys as well as 21+ year old models
Description
FAQ
Comments (54)
Looking good! If you want, I offer you to train it for you. I have a 3090 that's on idle most of the day. I just don't have time to make more datasets. I also caption everything by hand. I have dreamtrainer2 as well as dreambooth extension and kohyaSS. I can use your training parameters if you want me to or try mines. I had some pretty decent results. I found that fine-tuning works much better than dreambooth in most cases. I can use the checkpoint of your choice or plain SD1.5. You can always extract a lora from a fine-tuned checkpoint.
Thanks for this offer but i stop and rework many many times on big models :D so i wait for a RTX A5000 card ^^
@SysDeep Why an A5000? The RTX has the same chip, more CUDA and is cheaper?
@2BlackChicken get one from a reddit user ^^
Thank you very much ! + 10 stars!
What ever happened to the Gorilla Breasts ?
Thanks for this. I like how its in one lora, not a bunch of seperate ones or TI's.
This Lora works very well, generally. There's a few issues, none of which are show stoppers, but should be taken into account. (see below)
1. It doesn't work for every diffusion, sometimes it works for 3 out of 5, which is fine.
2. Breast size with this lora can directly correlate to a girls age. It can be difficult to give an adult woman a-cup boobs. Either the Lora, or the model really wants to draw non-adults with smaller breast sizes. 3 out of 5 will be kids.
3. There is an upper limit. This Lora doesn't do so well with oversized, saggy boobs, but it can work in combination with others to provide such effects.
thanks and yeah its still in progress :D - To 2 this is the fault of Checkpoints/vae :D so i add child as negative ^^
exactly, an inherit bias of sd. that's probably why cup sizes of mature women are more on the bigger size, seems that sd learned that breasts grow bigger the older a women gets :)
Thank you for this, Great work!
having some issues inpainting with this new version - kept trying medium breast and kept coming up flat
with inpainting I have unfortunately no experience trained and tested only txt2img :(
my issue is, it adds lace or net under garments to the model. do you have the same issue?
Don't use medium, large, small, flat, etc.
Use breast and give it weight : (breast:1.5), or (breast:0.2) for example. Weighting (flat chest:X.X) also gives great control, often better than words. Works on any breast, waist, etc. size. But also for anything really. The more weight the bigger, or the more you'll have of it if it's an object.
@hansolocambo what about lora weight?
Like a phoenix from the ashes 🙌
Legend! Thanks for not giving up. One Lora to rule them all..
Out of curiosity, how were these trained? Did you crop breasts out or were they trained with entire bodies?
Full Body / Upper Body Images
it is trained with all body or half, as long as the object that is emphasized looks "OPPAI" xD then you just have to correct each trigger where the photo and trigger are adjusted
thank you
sorry for stupid ask, but i'm new on stable diffusion (i use easy diffusion): loaded the lora, promted the trigger in the negative promt, but i don't noticed any change.. how it's possible?
Listed Trigger Words (Prompts) dont in negative
Negative prompts are anything you don't want in your image, so Stable Diffusion tries to filter this out.
To my knowledge, you also need to put the following tag in your prompt (anywhere) <lora:Breasts_Helper_Trail_v2:0.6>
Like you I see no change when I don't put this. I don't know why it seems to work for the creator of the Lora in their demo images despite not using the tag.
Note: it should be <lora:{the-lora-file-name}:{weight}>
Pay attention as the file name seems messy with this lora as it does not follow the version naming. I just downloaded v1.1 and the file name referance a v2
Would you kindly add nipple piercings? Much appreciated.
could you add one more gigantic arealos like this
http://tb.sb-cd.com/t/5840282/5/8/w:800/t8-enh/iori-yuki-s-big-tits-areolas.jpg
,thank you.
thats wild brudda
I keep getting black spot nipples. Very centre is quite dark, nearly black. Any ideas?
Have this sometimes, too. Inpainting helps, just mask and prompt nipples that you'd like.
Its very Hard to do perfect Lora with so many Trigger Words - sry but tricksey6182 say it - inpaint help for errors - 2.0 will not better :D many many more tirggers - At time 4029 Images 81 tirggers ^^"
I'm getting a erro when traying to download , just happens with this lora , anyone else experiencing this same issue ? any mirror for this lora ?
It's most likely your browser. I was able to download it just fine on edge.
Tested on: Chrome, Firefox, Opera, Edge - work fine
It makes sense after downloading the file to check its hash using any utility that does this. The hashsum for verification is taken from this page from the details of a specific version.
This lora will often bring out random slabs of cut meat. I don't know why.
On one hand it's hilarious, the other hand isn't available for comment at this time.
areola commands only affect size of breast
Need to use low weights or faces change from those predominant in Checkpoints.
I dont know why but all my traindata has no faces xD
@SysDeep did you have any luck with version 2 with less bias on the rest of the body & Face?
@SysDeep my bad, must be from the checkpoints
I need nipple piercings, please. PLEASE. I can submit files.
aint no way ur that horny blud hahahhaa
Has funny interactions with dark skin tones - consider maybe training with some dark-skinned images too?
we need this for SDXL!!
Biggest issue I have right now is that it totally changes the face of a LORA I'm using. Even at weight of 0.6, it changes the face entirely. Not very good when it does this.
nice and well-intentioned attempt, but this lora merciless overwrites my models! Say my faces backgrounds and poses don't work anymore. Therefore, this Lora is unfortunately useless for me.
Agreed - it's REALLY heavy handed with faces and hair for me. It does a terrific job on its specific task, but it's tough to use if I care about anything else in the frame. I feel like the training dataset included too much detail outside of its criteria without enough variation per-trigger. ie., "small breast" gets interpreted as something like "small breast, brown hair, bright eye makeup, by a swimming pool" because there wasn't enough variation for the model to know we're only talking about the breasts.
Nope all my Caption only focus on Tits - Handmade, all images has no faces / pose ... swap for next version to my own trained model for it to fix that ^^"
@SysDeep feels like the lora affects the face and overall art style like mentioned
Sad to hear you aren't doing a V3, but hopefully someone can carry the torch forward with a future LORA. V2 is a great resource!
Details
Files
Breasts_Helper_Trail_v2.safetensors
Mirrors
Breasts_Helper_Trail_v2.safetensors
Breasts_Helper_Trail_v2.safetensors
Breasts_Helper_Trail_v2.safetensors
Breasts_Helper_Trail_v2.safetensors
irl_breasthelper.safetensors
Breasts_Helper_Trail_v2.safetensors
Breasts_Helper_Trail_v2.safetensors
Breasts_Helper_Trail_v2.safetensors
Breasts_Helper_Trail_v2.safetensors
Breasts_Helper_Trail_v2.safetensors
Breasts_Helper_Trail_v2.safetensors
Breasts_Helper_Trail_v2.safetensors
Breasts_Helper_Trail_v2.safetensors
Breasts_Helper_Trail_v2.safetensors
Breasts_Helper_Trail_v2.safetensors
Available On (2 platforms)
Same model published on other platforms. May have additional downloads or version variants.