WAN2.2 - Boob Flash LoRA v0.1

Introducing the Boob Flash LoRA model, designed to enhance your creative workflow with the WAN2.2 I2V model. This LoRA model is experimental and intended to produce a simple breast flashing effect, where the shirt or top is lifted to reveal the breasts. This is my very first published WAN video LoRA and I am still working to improve my skills in this area.

Model Details

The Boob Flash LoRA model is an Image-to-Video (I2V) LoRA, designed to work with the WAN2.2 I2V base model. It's been trained to produce a realistic boob flashing motion. In my experience, the base model has difficulty creating a consistent and realistic breast flashing motion, often resulting in clothing being ripped off or suddenly disappearing in ways that defy the laws of physics.

This LoRA was trained to produce a more constrained and realistic movement pattern that more closely reflects what people usually expect from a breast flashing scene.

Usage Guidelines

The sample videos were created using a LoRA strength of 0.9, but feel free to experiment with other values. This model was trained on the trigger phrase "flashing her boobs", for example:

She is flashing her boobs

While this can sometimes achieve the desired results on its own, the current version 0.1 has a tendency to perform an overhead shirt disrobing motion and often requires additional prompting to prevent this. The complete phrase I used to generate the example video collection is below:

The woman is flashing her boobs. She lifts up her shirt to expose her breasts and nipples to the viewer. She holds her shirt up under her chin to reveal her bare chest and large perky breasts.

This LoRA has been tested in combination with other breast LoRAs, such as WAN2.2 Breasts Helper, Torpedo Tits, Big Perky Breasts, and WAN Nipple Fix, all of which result in more nuanced and realistic results. And yes, each one of the models listed above is a T2V LoRA model.

From my experimentation, I can confirm that T2V LoRAs work very well with the WAN2.2 I2V workflow and even enhance it in several useful and important ways. I2V LoRAs are trained purely on static images and therefore must rely on activating the base model's pre-existing motion patterns. I2V LoRA training has very limited influence on motion as a result.

T2V LoRAs on the other hand are trained on video clips and therefore can be taught very precise and complex motion patterns that do not already exist in the base model. These motion patterns can add much more realism in the case of breast movement, for example, and improve the results significantly.

Tips and Tricks

• I recommend pairing this LoRA with any of the models above (both high and low noise) at 0.5 - 0.8 strength) to generally improve the results and alter the breast and nipple appearance as desired. You can also experiment with any other breast LoRAs to achieve different results. Keep in mind that T2V LoRAs can also be used in the I2V workflow.

• WAN2.2 Breasts Helper in particular has proven to be very helpful as its name suggests, however, it has a strong tendency to turn the subject to face the camera directly. If you don't always want the subject to turn to face the viewer, you may have to lower its strength, use a different LoRA, or add more specific prompting.

• This model is best used alone mainly if you want to avoid any undesired or unintended effects that have been trained into other models, such as altering the pose, appearance, or facial expression of the subject.

• Experiment with different prompts to explore different effects and please share your findings with the community!

Future Development

This LoRA model is currently experimental and may not always produce the intended results. I greatly appreciate any feedback you provide as I work to improve the model.

If you find this LoRA model useful, please consider supporting me on Patreon where I create NSFW art celebrating ebony beauty, share my models, workflows, and more. Your support enables me to keep improving and expanding my offerings. I sincerely appreciate your consideration!

NOTE: This model is NOT intended to be used on photos of real life individuals. Please use responsibly!