I was tired of other LoRAs without the areolas textures and details. Most of the nipples I got until now using Flux was some soulless plain areolas without the depth it requires.
Then I curated a mid-size dataset with the best breasts and areolas I could find today in the internet.
I'm also releasing individual LoRAs for some of the trigger words listed below:
Included in Genral Lora and also Stand-Alone files:
big areolas - Rounded medium to big areola size
dark nipples - Dark nipples with more contrast to the natural skin
ghost nipples - Pale nipples that appears slightly
long nipples - Nipples that you can bite easily
puffy nipples - Small breasts with the tip of them covered with a dense nipple
saggy breasts - breasts hanging a bit due to the gravity
small areolas - Areolas with a small radius
veiny areolas - Veins appearing under the skin
Stand-Alone file only:
banana tits - Similar to puffy, but focused on breast shape instead of nipples only
silicone tits - The spheres from the 90s. It focuses on shape, you can combine other nipples
empty tits - Similar to saggy, but empty
flat tits - Flat chested woman
oval areolas - When it's cold, the areola shape may change to oval. Good texture
bumpy areolas - When it's cold, the areola texture becomes bumpy. Good texture
wrinkled areolas - When it's cold big areolas becomes wrinkled! Good texture
You can combine more than one. Inpainting them to improve the result is a good option.
Use a strenght from 0.7 to 1.5, depending on your scenario. Negative weight put clothes LOL
Note below for the General LoRA only:
This one tries to include all datasets from the other individual LoRAs. Trigger words (not really required at all) above "may" have some effect when inpainting and guided. If they don't work no problem, it's just Flux trying to use the best nipple to fit the girl. This is why there are individual LoRAs for each type.
Additional trigger words only present in the General LoRA:
ginger nipples - Nipples from the Redheads
regular nipples - Normal regular nipples and areolas. In doubt? start here!
Bonus that I added in case the full body appears, to have something right down there:
hairy pussy - hairy pussy of a woman standing
shaved pussy - shaved pussy of a woman standing
Flux does what it wants LOL. I trained different concepts, but flux does not follow them as it did with SDXL. Flux seems doing a breast based on the overall girl appearance, but it's possible to achieve using inpainting and some prompt engineering.
**Some few times it generates an almost translucent horizontal pattern in the image, and I don't know why
Description
FAQ
Comments (17)
veiny areolas... my hero...
@Skullkid what checkpoint do you use?
For training you can find here:
https://civitai.com/articles/3921/this-is-how-i-train-loras-updated-with-flux
For generation, I use Flux Q8: https://civitai.com/models/647237?modelVersionId=724149
can anyone remove the plastic feeling of flux ???
no lora seems to be able to achieve realism.
Boring reality, Grainscape, 200 Analog Core, Samsung Ultrareal. Using some of these helps. Using al person loras helps even better.
Yes
change the checkpoint, try "ultrareal fine tune"
https://civitai.com/models/952543/detailed-imperfect-skin-faces-and-torso-for-flux this one is good imo
https://huggingface.co/vafipas663/flux-krea-extracted-lora/tree/main
Use this lora at0.5 strenght
@Skullkid sorry bro small question, it says flux krea..does it work with flux1 dev?
@napsumfur225 the link above is the extracted Lora from Krea to use on Flux dev, so it adds the difference that krea has related to flux 1. This improves flux dev output a bit.
@Skullkid thank you so much man you are awesome
I think this Lora was undertrained on a limited dataset. I have generated about hundreds of images with trigger words like "long nipples" and "big areola" etc. Every single time there is a specific pattern of tan on breasts. The shape is also saggy almost always. Thank you for the Lora anyway but I just said what it felt.
Could you tell me what tool did you use for training and how many images did you train it on?
Hello. You are correct. Due to the amount of breast types I did train, I limited to 25 images of each type. Note that I had to find different angles, skin colors, shapes and illumination for each type, what requires some work.
I had to add also some full body images of multiple naked woman to avoid having the Lora creating breasts on wrong body places.
At the end, it's a bit limited, I understand you concern. But Flux lora with multiple concepts is not too good as it was on SDXL, and then I had to create the individual loras for each breast type.
The total amount of images is 400, I used kohya SS. I tried multiple times with multiple dataset configurations until I got a good result
"Negative weight put clothes LOL"
LOL!