
Support my work on Patreon and Ko-Fi and get access to tutorials and exclusive models.
Rate and leave a like if you enjoyed it, and follow for new models!
LoRA trained to generate a character covering their breasts and/or their crotch. I used many different tags in order to achieve multiple angles and with various poses. It is very flexible and does not affect style or composition.
By default, using the lora without triggers will usually already have some effects.
How to use:
covering breasts (default action, can result in breasts covered by one arm/hand or both;
covering crotch (can be combined with covering breasts)
crossed arms (should be combined with covering breasts for better effect)
hands on own breasts (each hand covering each breast), or for similar effect you can use hands on own chest, or breast press.
you can combine the above with a tag to change the perspective from behind the character, e.g.: from behind, looking back, ass, etc.
when using covering breasts, if you want to use only one hand/arm to cover the breasts you can describe the other arm's pose, e.g.: arm at side, arm up, hand on hip, arm support, hand on own face, finger to mouth, etc.
other tags that can modulate the effect: breasts, nipples, cleavage, etc.
To keep the character with clothes on you should describe them in the prompt, but you can also add some tags in the negative prompt, e.g.: nude, naked, nsfw, breasts out, etc.
Description
FAQ
Comments (18)
Havent tested compatibiity with other cloth/character lora's but if its compatible with those too this is a 10/10 LORA
thanks! it actually works with multiple loras. I tested it together with a couple of my other loras, a character lora and a style lora, at the same time.
The Master LoRAist returns with another instant classic! I really love the themes you share with us.
Finger issue :(
EDIT: Inpainting solve it!
hey there, as you can see by the many examples here, there are limited noticeable fingers issues. I haven't seen that many issues when doing my testing. much less than usual.
the hands/fingers issue you may be experiencing is not due to the lora, but most likely due to the base model you're using. however, in any case, it's the biggest problem in SD in general, that no SD1.5 base model or lora can solve.
Oh hey if you're talking about extra fingers appearing on one of the breasts when you have just one arm covering, adding "crossed arms" or even "hands on own breasts" to the negative prompt should get rid of that issue mostly. I had a few of those issues but that solved them for me :)
@Phantasmastic thanks, good to know. it may be due to the training of all of these poses together so adding some of the trained triggers in the negative should help.
@psoft Yeah it's definitely just because the lora includes all the poses, I have generation previews on in my webui and I've noticed that on very early steps it can generate people with like four arms, two covering the breasts from either direction, and two for the crotch, it just doesn't get rid of one of those extra hands on the breasts later down the line unless you add the negative or, as OP noted, just inpaint :)
Is this compatible with DiffusionBee? I get an error that says "module 'numpy' has no attribute 'bfloat16'". I'm new to LoRAs, so I might be doing something wrong.
I haven’t checked diffusionbee, but this is a basic lora, so it should work if diffusionbee is compatible with loras. Do other loras work for you?
Amazing!
How are you able to control the effects so well?
I admire your outstanding work!
I tryed another older LORA for "covering herself" and it was terrible, introducing quality drop and body distortion.
This one is much better about this point, but I notice a bit of quality drop on the image overall (not distortions).
The real big problem I got with this LORA is it keeps to render vagina distorted, like there was a buldge.
So the result is aceptable only if the girl cover herself crotch completely...
Since is the 0.1 version I hope there will be improvements in future.
that's odd. does it happen in a base model in specific or in general?
@psoft
I was experimenting with a checkpoint called "RealMoon v8.0".
after some other experimens, I see it tends to adultifize the body. AKA probably the training images were all about well curvaceues women.
Also, I noticed if I lock the seed, NOT using the LORA and then using it, it will change completely the resulting image. I suspect there still are too few trained images in the data set.
@settima_ai yeah, I suspected it could be the base model. some merged models can get weird results with some loras. this lora should not change the image that much, especially with a sampler like Euler A.
for reference, this lora was trained on ~1150 images
@psoft really a lot! O_o
I think i know which you are referring and it seems this one too merges with others when there are multiple characters in the scene
好好好
Details
Files
covering_breasts_v0.1.safetensors
Mirrors
covering_breasts_v0.1.safetensors
covering_breasts_v0.1.safetensors
C_covering_breasts.safetensors
covering_breasts_v0.1.safetensors
covering_breasts_v0.1.safetensors
covering_breasts_v0.1.safetensors
covering_breasts_v0.1.safetensors
covering_breasts_v0.1.safetensors
Covering_breasts_crotch.safetensors
covering_breasts_v0.1.safetensors
covering.safetensors
covering_breasts_v0.1.safetensors
covering.safetensors
covering_breasts_v0.1.safetensors
covering_breasts_v0.1.safetensors
covering_breasts_v0.1.safetensors
Available On (2 platforms)
Same model published on other platforms. May have additional downloads or version variants.

















