LoRA trained on videos with women lifting or pulling down their shirts to reveal their breasts.You should probably leave the strength at 1.00.For promting, you should first describe the beginning of the scene:<pre><code>This digitally rendered animation shows a beautiful young woman wearing a red dress. She is looking behind her, with her hands over her breasts.</code></pre>Then, you should describe the actions she takes one after the other, with reference with how much time elapses:<pre><code>She starts the video shyly looking at the viewer and pulling her dress down. One second later, her breasts are revealed. Then, she pinches her nipples. Finally, she pulls her dress back up.</code></pre>Finally, add booru like tags with all the elements you want to appear in the video:<pre><code>breast drop, clothes pull, breasts, lifting own clothes, undressing, no bra, dress, nipples</code></pre>Look at the example videos to get more examples.Fair warning, this LoRA has more misses than hits. I recommend generating a low resolution video first and then upscale the promising ones, like in the examples workflow.<h1 id="v2-luocpstke">V2</h1>When I trained the first version I realized that most of my training examples weren't used. I adjusted the options and retrained with more of my data. I think the results are a bit better. At the very least, the boobs are jigglier :P.

Files with SHA256: 320f3f75e86528ca68d2614e04f196fa9850fa39b8489dd4d9b11a1ed64a88c4