This LoRa helps to depict beautifully shaped breasts in various sizes.

The LoRa was trained using 400 carefully selected close-up images.

When character LoRa is used, it does not change the character itself (except, of course, for breast size).

The triggers used were "small breasts," "normal breasts," "medium breasts," "large breasts," and "extra large breasts." Without a trigger, breast size depends on the checkpoint and other prompts, but even in this case, using LoRa results in a significantly better representation. (See the girl with pink hair.)

A weight of 0.7 generally yields good results. For "large breasts" and "extra large breasts," the weight can be reduced to 0.5.

The example images are as follows:

- No character LoRa was used for the girl with blonde hair.

- A character LoRa was used for the girl with dark hair, which was trained without captions. Therefore, breast size was not explicitly trained.

- For the girl with red hair, a character LoRa was used, which was trained with image captions. This explicitly included training her breast size.

- For the girl with pink hair, no character LoRa was used, and no trigger was specified.