Premium character LoRAs
Original synthetic characters, bundles and AI tools are in my shop:
Join the community on Discord: https://discord.gg/nB79wRNtMT
Back Dimples
A concept LoRA for Krea 2 focused on back dimples (dimples of Venus): the lower back, waist and buttocks seen from behind, in thongs, bikini bottoms, low-rise jeans or nude. It gives an amateur photo look.
Usage
Start with the trigger word BackDimplesRK, then describe the two dimples above her buttocks (how deep and visible), her lower back, waist and buttocks, what she wears or that she is nude, her pose from behind and the place, and end with the light. A side light that shapes the dimples brings out the effect the most.
BackDimplesRK, Photograph of a woman seen from behind, with <hair>, in <a thong / low-rise jeans / nude>, two deep dimples on her lower back, <pose> in a <room>. The lighting is <side light / light>.
Recommended weight: 1. Made for Krea 2 Turbo (8 steps, CFG 1).
Prompt Examples
"BackDimplesRK, Photograph of a nude woman with a medium brown skin tone and wet, dark brown, shoulder-length hair. She has a slim, toned physique with visible dimples of Venus on her lower back, above her buttocks. She is fully nude, kneeling with her back to the camera, showcasing her buttocks and the curve of her hips."
"BackDimplesRK, Photograph of a nude woman with light skin and long, straight blonde hair lying face down on a white bed. She is fully nude, exposing her back, buttocks, and lower back tattoo on her right side. Her buttocks are prominent and she has visible dimples of Venus on her lower back."
"BackDimplesRK, Photograph of a nude woman with light skin and brown hair, styled in a single braid. She is on all fours on a grassy ground, with her back arched and her buttocks raised. Her lower back shows two prominent dimples of Venus."
Training
- Base model: Krea 2
- Type: Concept LoRA
- Steps: 4000
- Rank: 32
- Optimizer: adamw, constant learning rate
- Engine: ai-toolkit
Generation settings
Trained on Krea 2, and works on Krea 2 Turbo for faster generation.
- Raw: ~30 steps, guidance 3.5
- Turbo: ~8 steps, guidance 1.0
The gallery images were generated with krea2_turbo_mxfp8.safetensors, at 8 steps, guidance 1, LoRA weight 1.