Premium character LoRAs
Original synthetic characters, bundles and AI tools are in my shop:
Join the community on Discord: https://discord.gg/nB79wRNtMT
Cum On Tongue
A concept LoRA for Krea 2 focused on cum on the tongue: open mouths, tongues out, cum on the tongue, lips and chin, eyes looking up at the camera, in close-ups and shots from above. It gives an amateur photo look.
Usage
Start with the trigger word CumTongueRK, then describe her open mouth and tongue, how much cum and where (tongue, lips, chin), her expression and eyes, her skin, hair and makeup, what she wears or that she is nude and the framing, and end with the light. The more precise the mouth and the cum, the stronger the effect.
CumTongueRK, Photograph of a woman with <hair>, mouth open and tongue out with cum on <where>, looking up at the camera, wearing <... / nude>, <framing> <place>. The lighting is <light>.
Recommended weight: 1. Made for Krea 2 Turbo (8 steps, CFG 1).
Prompt Examples
"CumTongueRK, Photograph of a topless woman with fair skin and shoulder-length brown hair. Her eyes are half-closed, and she is looking up at the camera with her tongue sticking out, covered in a significant amount of white cum."
"CumTongueRK, Photograph of a woman with fair skin and light brown hair, tied back. Her mouth is open, tongue out, and she is performing oral sex on a man. Her tongue is covered in cum, and there is cum on her lips and chin."
"CumTongueRK, Photograph of a nude woman with medium brown skin and long black hair. She is kneeling on a tiled floor, her head tilted back, with her mouth open and tongue extended, receiving cum from a man's erect, uncircumcised penis. The setting is a room with white and orange tiles, and the lighting is bright, highlighting her skin and the cum on her tongue."
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.