UPDATE - added LoRA (v3.1)After many requests I decided to try a LoRA again for this concept. Increased the dataset, removed poor quality images or too diverse, reworked captions, removed token.Because this is a complex concept, I opted for not using a unique trigger word. Instead you can should typical tags you'd use for this type of concept, but the effect does vary depending on the base model. It takes a bit more work when prompting than my usual models. Please let me know how it works for you and the base model you use.Here are some examples, but see also my image gallery:<ul><li>male tags: 1boy, 2boys, ..., multiple boys</li><li>grabbing tags: grabbing, breast grab, ass grab, crotch grab, torso grab, grabbing another's ass, arm grab, etc.</li><li>undressing tags: skirt lift, shirt lift, shirt pull, etc.</li><li>other tags: fingering, touching, masturbation, etc.</li></ul><hr /><s>LyCORIS/Locon version (1.5i-locon), old version</s><s>This is a LyCORIS/Locon model aimed at preserving your base prompt and then add the concept of groping, molestation, chikan on top, without affecting style and with small impact on composition. I always try to aim at having my models not overfitting so that they're more flexible to your prompt and allow you to use different base models together with LoRAs and TI embeddings without burning the image, which is why it took me &gt;30 runs to get this one right...</s><s>It can be used from multiple angles, with whatever character description, and it can be used in combination with your favorite character LoRA, as well as style LoRAs.</s><s>You can use weight 1.0 or above, with your usual CFG.</s><s>This model does not require any trigger word, as it was trained without one, but you can use the following tags to increase/modify the effect, besides increasing the weight beyond 1.0: </s><s>multiple boys, groping, grabbing, grabbing from behind, breast grab, ass grab, molestation, chikan</s><s>, etc. You can use many more related tags since the model will amplify their effect.</s><s>There is a bias that the location is a train interior since it was prevalent in the chikan dataset, but you can prompt other backgrounds as well.</s>Additional infoI make these models for fun and for the challenge and release them for free for everyone to enjoy, for non-commercial use (see license). The one thing I ask of you is to provide rating and feedback! That helps me understand where I can improve and helps me stay motivated to produce more content!I accept suggestions or requests here or on Civitai discord server.You can also check my more polished images and upcoming models on my <a target="_blank" rel="ugc" href="https://www.pixiv.net/en/users/25545070">pixiv </a>profile.

Files with SHA256: 84c5e8d63a0a81c7b04f9167f948d114f3bb993e13a116e3cea69aa33fd0b395