NEW VERSION: Flat Clamps (V1.0)
Trained on images of just clothes pins to prevent bleeding between the different types. Works better but this one is probably the most finicky of all of them so far.
Primary trigger: clothes_pins
NEW VERSION: Flat Clamps (V1.0)
Trained on images of just flat nipple clamps to prevent bleeding between the different types. Works much better.
Primary trigger: hyvid_nip_clamps, flat_clamps
NEW VERSION: Standard Clamps (V1.0)
Trained on images of just standard nipple clamps to prevent bleeding between the different types. Works much better.
Primary trigger: hyvid_nip_clamps, standard_clamps
NEW VERSION: Clover Clamps (V1.0)
Trained on images of just clover clamps to prevent bleeding between the different types. Works much better.
Primary trigger: hyvid_nip_clamps
A LoRA designed to generate various types of nipple clamps and nipple stretching. Trained on 200 images, 30 Epochs, 6030 steps
This first training is a bit rough, definitely gives some body horror outputs fairly often. Biggest issue I believe is I included too many clamp types and concepts in a single dataset causing the model to overlap and confuse them. V2 will be training each clamp type individually.
Primary trigger: hyvid_nip_clamps
Clamp Type Triggers: standard clamps, flat clamps, clover clamps, pinch clamps, clothes pins
Clamp Modifiers: chain on clamps, weights on clamps, weights on chain, chain attached to collar, chain in mouth, chain attached to gag, hand holding chain,
Nipple Modifiers: pulled nipples, nipples pulled down, nipples pulled up, nipples pulled out
Clothing / Restraint Concepts: ball gag, leather collar, metal collar, rope around (neck|breasts|stomach|waist|legs), wrists bound, blindfolded, leather harness
Poses: arms bound behind head, hands behind back, laying on back, sitting, kneeling, standing, eyes closed, leaning over, legs spread
Other Concepts: pain face, drooling, (small|medium|large|huge) breasts, bdsm dungeon background
Views: close up, side view, face partially out of frame, head out of frame, low angle shot, high angle shot
As you can see I probably tried to do too much.
All Examples videos are purely using the original training captions which you can see here. (In wildcard prompt format)
https://files.catbox.moe/3s9n9a.txt
Description
Trained on images of just clothes pins to prevent bleeding between the different types. Works better but this one is probably the most finicky of all of them so far.
Primary trigger: clothes_pins