AI porn models for adult image and video generation
Browse the building blocks behind the factory. Filter checkpoints and LoRAs by image or video support, base model, and tags, then open any version directly in the unified generator.
Masturbation Cumshot for Wan I2V 14B
This LoRA model is designed for the Wan 14B I2V model to generate realistic cumshot scenes. It allows some control over where the cum lands using specific prompts and produces more realistic results on bodies and objects. The model was trained with improved datasets and captions at a higher resolution.
Ahegao Face Wan (i2v)
This LoRA model generates ahegao faces, typically including drool and tongue protrusion. It was trained on a small dataset of videos and images and may sometimes produce sharp, vampire-like teeth. Users can try mitigating this by adding "sharp teeth" and "vampire teeth" to the negative prompt.
Wan 2.1 - Erect Penis, Cock, Dick, Lora, I2V
My other penis Loras can't do a penis facing more toward the front, so here's one that can. Main keyword is "penis" and it's also useful to say "facing the camera" or "facing to the side" "slightly to the side" to get different angles.
Wan 2.1 Big Natural Breasts - Saggy Tits, Huge Breasts, Lora, T2V
This Lora enhances the Wan 2.1 model to generate images and videos featuring larger, natural-looking breasts. It's trained on the 14B model and suitable for both image-to-video and text-to-video applications; version 2 includes video data in training. A negative prompt addressing long fingernails may be needed.
Wan 2.1 Deepthroat / Facefuck Entry - Blowjob, Fellatio, Oral, Lora, T2V 14B
This Lora generates videos of blowjobs, often starting with the penis outside the mouth and leaning towards facefucking. Video quality can be inconsistent but improves when used with other Loras at lower strengths, requiring multiple generations for optimal results. The action and details are generally good.
How checkpoints and LoRAs fit together
Start with a checkpoint
A checkpoint controls the base rendering system. Its supported modes determine whether it can create images, edit an input, or produce video.
Match the base model
A LoRA must target the same base-model family as the checkpoint. Similar names do not guarantee compatible weights.
Check the validated workflows
A catalog version can be documented before it is generator-ready. Only versions with validated broker and artifact mappings can spend credits.