Wan 1.3b NSFW is a NSFW full fine tune of the Wan 1.3b model. Updated to an even better performing i2v model. Designed to get as close to the 14b as possible for NSFW.
The model can do missionary, doggystyle, cowgirl, reverse cowgirl, blowjob, handjob and should be able to do anal and titfuck as well. Should work with both straight and gay/bisexual generations as well!
Works on 8GB of VRAM! Great if "GPU poor", and a lot faster than the 14b.
Works with 2d, 2.5d, 3d, furry and realistic!
This is a i2v model but should work on t2v as well.
The model was trained on a large variety of NSFW terms so specifying what the generation is supposed to be about (i.e gay, bisexual, hentai, yiff, realistic) should help get the desired results.
PROMPTING GUIDE
↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓
Prompting for the motion you want helps using terms like "smooth motion" will help since this was trained into the model.
Always experiment with your prompting for best results
Prompt example
A 2.5D animation depicts a scene from a first-person perspective. A nude woman with black hair is positioned beneath a man having sex in the missionary position. The man’s forceful, rhythmic thrusts drive his hips into her, each movement causing his penis to slide completely in and out of her vagina. Her breasts, soft and supple, bounce with his motions, their curves stretching and contracting with every thrust. The warm, golden light of the room bathes the scene, casting gentle shadows that trace the contours of skin and highlight the smoothness of her body. Her movements are captured in exquisite detail, with every bounce and shift of her breasts rendered in fluid, dynamic motion. The background is softly blurred, its edges dissolving into a hazy glow that directs focus to the intimate, lifelike interaction between them. The animation maintains a high frame rate, ensuring every nuance of her movement is preserved—her skin shimmers subtly under the light, and the interplay of motion and texture creates a seamless, immersive flow. The scene is rendered with meticulous 2.5D depth, emphasizing the realism of their physical connection and the natural rhythm of the moment.
Negative prompt example
overexposure, static, blurred details, subtitles, paintings, pictures, still, overall gray, worst quality, low quality, jpeg compression residue, ugly, mutilated, redundant fingers, poorly painted hands, poorly painted faces, deformed, disfigured, deformed limbs, fused fingers, cluttered background, three legs, upside down, white flashlights, glow light effect, post edit, ugly distorted hands, scrambled fingers, wild arm movement, text, icons, logo, post production, a lot of people in the background, motion blur, camera movement, impossible anatomy, stuttering, rough motion
V2
Updated using the ODE i2v causal model checkpoint from the causvid team, the model now performs alot better on i2v. ODE i2v checkpoint is a pure quality/motion and prompt adherence enhancement with the same speed as before, it is not the causvid one that decreases quality for increased speed
V1
It was created by merging a full NSFW 1.3b finetune with the wan fun 1.3b inp v1.1 model combined with multiple quality enhancing LoRas+NSFW motion LoRas. Combining it with the wan fun model enhanced the quality significantly due to the fun model's better architecture.
This model should be able to be used as a base model to train LoRas with to achieve better NSFW LoRa creation since the model is already finetuned on NSFW
All characters are 18+. This is intended for mature audiences only you and you alone are responsible for what you generate. THIS MODEL IS NOT TO BE USED ON ANY REAL PERSON! 18+ ONLY!