This LoRA is based on the Coyote Jocks Cotton Ringer Collection, including a shirt, tank top, briefs, jockstrap and socks in the colors black, blue, and red.

You can try for colors outside of the training set, but YMMV. See the "Version Details" for details and limitations. If the model confuses jockstraps with briefs or vice-versa try putting the one you don't want in the negative prompt.

Trigger Words

The dataset included the three official color variations (black, red, and blue) as well as artificially created images with different colors. The hope is that the model can generalize beyond just a few colors:

• shirt: white shirt with <color> trim

• tank top: white tank top with <color> trim

• briefs: white briefs with <color> trim

• jockstrap: jockstrap, jockstrap with white straps, jockstrap with <color> waistband

  • The model should understand jockstrap to create a plain jockstrap, jockstrap with white straps to include the straps, and jockstrap with <color> waistband to include the waistband.

  • For SD 1.5 (V1) they were tagged white jockstrap with <color> waistband

• socks: white socks with <color> stripes

Photo types:

• professional studio promotional images were tagged studio shot

• amateur photos were tagged casual photo

Versions

PONY (V1)

Use weights 0.8 to 1. The model is pretty good at re-creating the shirt, tank top, briefs, and socks. Pony has a good understanding of what a jockstrap is so it re-creates them more faithfully than SD 1.5.

SD 1.5 (V1)

Use weights 0.6 to 1. The LoRA is good at re-creating the shirt and tank top. It's ok at re-creating the briefs. It struggles with the jockstrap. It can re-create the front of the jockstrap, but not so much the back straps.