Check out my websites for more tentacle/vore galleries:<ul><li><a target="_blank" rel="ugc" href="https://subscribestar.adult/artificialvore">https://subscribestar.adult/artificialvore</a></li><li><a target="_blank" rel="ugc" href="https://www.deviantart.com/tentaclevoreai">https://www.deviantart.com/tentaclevoreai</a></li><li><a target="_blank" rel="ugc" href="https://www.pixiv.net/en/users/112576191">https://www.pixiv.net/en/users/112576191</a></li></ul>My first attempt at SDXL LoRA!I'm a huge fan of Lovecraftian monster vore on human girls, and though there are some awesome LoRA like <a target="_blank" rel="ugc" href="https://civarchive.com/models/429795/worm-vore">worm vore, </a>I wasn't fully satisfied with the results and really wanted to train a model of my own. Since I couldn't train SDXL on my PC, I used CivitAI's cloud service to generate this model.The main prompt is really just "vore". Combining too many prompts, or other concept LoRA will produce less vore-ish images, but don't let that stop you from being adventurous!I tried adding tags like "upper/lower body swallowed" "head/feet first" to sample images when training the model but they don't seem to do much. I even feel like those tags confuse the AI. Still, you could try them at your own discretion.<ol><li>Head first/Upper body swallowed I made <a target="_blank" rel="ugc" href="https://civarchive.com/models/845735">a new LoRA </a>specifically for this purpose so please check it out! Although I trained quite a few images for this concept, I found it difficult generating what I wanted. I tried "upper body", "head" for negative prompts but they didn't really work for me. Try using as fewer prompts as possible and use prompts like "ass", "legs", "feet", and don't use tags that define upper body, or head or face; with some luck you will generate very interesting images!</li><li>Feet first/Lower body swallowed The AI produced satisfying images most of the time, even with several other LoRA and many descriptive prompts. I used "lower body swallowed", "feet first" prompts but I think it works just fine without them. </li></ol>SDXL is so awesome! I thought AI was underwhelming when using SD 1.5, but it's a huge improvement over the previous version, and to think that SDXL is kind of outdated right now is very exciting. I can't wait for the future models to produce more precise results that I want :)

Files with SHA256: a1aa5afaf08d7bf53f0c545e94321c659f0d8d42f709313556cbf5898e120c3d