This LoRA is designed to improve the visibility and appearance of the chin in POV shots of the 69 position, which is something that base_v1 tends to struggle with.
However, the POV composition where the woman is positioned upside down on top of the man has not been fully learned. As a result, the woman tends to sit on the man's crotch even when "vaginal" is not specified. This is also an issue with my POV Chest LoRA, and remains a challenge for future development.
For v0.0.0, I used a total of 55 training images: 46 standard training images and 9 additional images with compositions closely resembling the 69 position.
Description
Initial.
dim8/alpha8, AdamW8bit, UNet LR 1e-04, constant 600 + cosine 150(150/200), 512² resolution, 55 images.











