This is a pure motion Lora. It is for use only on stylized fictional/generated adult characters. It does not interact with clothing. It contains no reference to ???? state or clothed state, and ignores any ???? or dressed state changes. It contains no ????? ??????? tagging outside of "ass". It is trained on cropped un-identifiable female torsos in motion. It may be ????, but that is purely decided by the user; start image, user intent, and user prompting. The user is responsible for their own misuse. I'm tagging it ???? anyways because the motion is pretty ??????. This injects 81 frames (5 seconds) of half-dancing, open-handed self-caressing, and hypnotic hip/torso sways and grinds. Entirely female focused. All my preview start-images are generated fictional characters. Metadata can be provided. Hopefully that's clear enough.

Master caption:

She is doing a ???? move.

Movement genres:

Gyrating her ???? in a circular hula motion.

Grinding forward and back with her pelvis repeatedly.

Wiggle dance.

Feeling her back and ??? with her hands. From Behind.

Dipping her ???? backwards towards the camera while she is posed ????????????. From Behind.

Feeling her body with her hands.

Posing sexily.

Squeezing her ??????? together.

Rubbing her breasts.

Turns around.

Bending forward.

None of these are fully needed; if you wish to pursue a sub motion like grinding, you can just use only that phrase on it's own. Doesn't have to use female pronouns. It will probably feminize whatever. Some of those overlap base Wan2.2 motion, which this was mostly developed to enhance and also let the base model shine through it. Run this at 0.8-1.0 to conquer it. Just use a spicy prompt and the start image will mostly inform the motion.

The best turbo lora motion comes with 6 steps, 3-3 split and 7.0 shift. With 4 steps: you may get better results with a High 3-1 Low split. This is trained on large 512 and 768 square motion cropping. Wan2.2 can properly scale it's flow motions about 1 bucket, it seems. The smallest output size you'd want is probably 448 pixel width, the largest pixel width is probably 1024. I'm not sure but it seems like that's the case.

Why'd I train this? Just kind of dissatisfied with existing ones in the grinding and hip movement genre—they often required stacking 3-4 LoRAs together to achieve this desired effect. Many people don't realize it's possible to pack multiple motion concepts into a single LoRA with the proper setup. My goal was to develop a versatile "secret sauce" for ???? motions that works across most poses and compositions, ideal for when you want something zesty but lack specific ideas.

I haven't extensively tested it with other LoRAs, but there are likely some good combinations out there—though layering might lead to some odd results. This LoRA has no influence on faces, facial expressions, or any motions beyond what's described in the captions. You'll always have better results with further prompting and embellishment.

Want more? 1$ is 100 steps on another Lora: https://ko-fi.com/tenstrip

My next Lora is a Female POV/Receiver ??? motion Lora. Already funded.