Don't do many images including this part of the female anatomy, but don't like the rubbery results from Moss.
Been playing with Moss LoRa training and must say, the model/checkpoint makes a huge difference in the results when generating with the LoRa! The base model and a couple of merged checkpoints give great results, others overpower this LoRa too much.
The samples are using either Moss Base INT8 or the excellent Realism By Stable Yogi checkpoint.
The strength will affect the labia size, but the texture/shape is retained at lower strengths.
If any additional descriptors are used in the prompt, the LoRa responds well. Prompt "huge" and you will get...well let's say something quite large. Samples had no descriptors added, just prompted the word "labia".
Works with various skin tones, builds, breast sizes, and will not affect faces.
Please post your generations here. Love to see your work, let me know what you think!
Description
Initial release



