Wan 2.2:

This model was trained using ai-toolkit with 20 81-frame 16-fps videos.

Wan 2.1:

Optional: You can optionally use the "NEGATIVE" version of this lora (with a negative value around -0.5) to that can help combat failed generations like her penis disappearing.

Positive Prompt Example:

Trans woman laying down. Thrusting all the way in and out of her asshole. Her penis is erect. Anal penetration. The camera stays stationary. She has her eyes open.

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 limbs, fused fingers, cluttered background, three legs, logo, text on screen, vagina, deformed penis, motion blur, blurry penis, eyes closed

This model was trained with 37 2-second videos at 16fps.

I'm aware that it's the general convention to train on the T2V Wan checkpoint, but I get bad results (with both T2V and I2V) when I do so with this particular dataset.