<h1 id="okay-just-hear-me-out">Okay, just hear me out</h1><h3 id="if-someone-controls-tentacles-then-this-someone-should-be-able-to-use-tentacles-as-clothes...-right">IF someone controls tentacles THEN this someone should be able to use tentacles as clothes… RIGHT?</h3><hr /><h1 id="description">Description📜</h1><hr />This LoRA allow you to place tentacles on character's body without tentacles in air, on floor, <s>in ass</s>, etc. And it can also cover things that normal clothes usually hide.There are LoRA versions for Illustrious and for Anima. Look below and find the description you need.V1.0 for Anima and Illustrious was trained on same images and provide simillar results, but require different prompts.<hr /><hr /><h1 id="anima-v1.0">Anima V1.0</h1><hr /><h1 id="prompt">Prompt ✍</h1><hr />Anima requires a prompt that is different from Illustrious, so the prompt has been changed.<h3 id="trained-phrases">Trained phrases 🏋️‍♂️</h3>LoRA doesn’t require any main tag; just use this phrases straight away. I think their meaning is clear without explanation, so I’ll only include comments where they might be needed.<ul><li>tentacles around hands</li><li>tentacles around legs</li><li>tentacles around breasts - usually don't cover nipples</li><li>tentacle cover nipples - can be used with or without previous phrase</li><li>tentacles around waist</li><li>tentacles on crotch - should hide genitals</li></ul>Also you can add tentacle's color in beginning. Example: blue tentacles around breasts. Or you can also try to use {color} tentacles separatedly from other phrases, but colors may go wrong.<h3 id="additional-phrases">Additional phrases 🛌</h3>There are also a few phrases that weren’t in the dataset, but they still work well.<ul><li>tentacle over eyes</li><li>tentacle around neck</li><li>tentacle bondage</li><li>tentacles all over the body</li></ul>Also, LoRA works well with other phrases that not from dataset. So you can try using your own phrases. Try to use tentacles around {something}.<hr /><h2 id="compatibility">Compatibility</h2><hr /><span style="background-color:rgb(26, 27, 30);font-family:-apple-system, BlinkMacSystemFont, &quot;Segoe UI&quot;, Roboto, Helvetica, Arial, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;;font-size:16px">This LoRA was trained with<span style="background-color:rgb(26, 27, 30);color:rgb(193, 194, 197);font-family:-apple-system, BlinkMacSystemFont, &quot;Segoe UI&quot;, Roboto, Helvetica, Arial, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;;font-size:16px"> <a target="_blank" rel="ugc" href="https://civarchive.com/models/2458426/anima?modelVersionId=2945208">Anima base v1.0</a>. I tested LoRA with it and with <a target="_blank" rel="ugc" href="https://civarchive.com/models/2642932/janima?modelVersionId=2967640">JANIMA v1.0</a>. Both works good. I also tried few character LoRAs and there was no serious problems.Maybe, here is just one small issue with this LoRA. It seems that it may have a slight impact on the style. But I am not sure.<hr /><hr /><hr /><h1 id="illustrious-v1.0">Illustrious V1.0</h1><hr /><h1 id="how-to-use-it">How to use it🤔</h1><hr /><h3 id="strength">Strength💪</h3>This LoRA has minor difficulties with its strength, but I recommend to try with the following strengths:<ul><li>At strength=1.0, LoRA will follow the prompt more accurately, but the background may become blurry.</li><li>At strength=0.8, LoRA will follow the prompt a little less accurately, but the background will be fixed and there will be minor issues. Personally, I prefer this value.</li><li>At strength=0.7 or lower, accuracy decreases and LoRA starts adding tentacles that do not lie on the body. The lower the value, the higher the probability of this happening. But you can use it, if you like it. It's not a bug, it's a feature.</li></ul><h3 id="main-tags">Main tags✍</h3>LoRA requires two tags: tentacles and tntcl_outfit. These two tags are sufficient to accomplish something. However, you will most likely want to control where the tentacles will be located. To do this, you should use the control tags described below. When using multiple control tags, one of the main tags can be removed.<h3 id="control-tags">Control tags🕹</h3><ul><li>tntcl_breasts - places tentacles on or around the breasts, often covering the nipples. To prevent the nipples from being covered, simply add the tag ‘nipples’ to the positive prompt.</li><li>tntcl_cover_nipples - places fewer tentacles than the previous tag, but places them so that they cover the nipples. Sometimes it makes the tentacles semi-transparent, which makes the nipples visible. To prevent this, add the ‘nipples’ tag to the negative prompt.</li><li>tntcl_crotch - usually makes two (less often one) tentacles that cover the crotch. If you use only this tag and the main tags, other tentacles may often appear. I haven't figured out how to deal with this, but in combination with other control tags, it works better.</li><li>tntcl_waist - places few tentacles around the waist.</li><li>tntcl_legs - places few tentacles around the legs.</li><li>tntcl_hands - places few tentacles around the hands.</li><li>{colour} tentacles - determines the colour of the tentacles. Works with all colours I have tested (I tested many of them, even rainbow one).</li><li>tentacle over {something} - sometimes allows you to add a tentacle to a specific place. For example, on the shoulder. LoRA was not trained to implement this feature. It works, but not always.</li></ul><hr /><h2 id="compatibility">Compatibility</h2><hr /><span style="background-color:rgb(26, 27, 30);color:rgb(193, 194, 197);font-family:-apple-system, BlinkMacSystemFont, &quot;Segoe UI&quot;, Roboto, Helvetica, Arial, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;;font-size:16px">This LoRA was trained with WAI-illustrious-SDXL v16. I tested about 10 different checkpoints, and LoRA works well with all of them. Some other checkpoints are still require change the LoRA's strength.<hr /><h2 id="how-it-was-made">How it was made</h2><hr />First, I created a dataset of about 70 images. I generated the images as usual, and then manually added tentacles using sketch inpaint feature. Then, based on this dataset, I trained the first LoRA. This LoRA worked in a similar way as this one, but it was too random due to the large number of tags and small number of images. However, this LoRA allowed me to create a larger dataset. When creating the new dataset, I took into account the problems of the previous one. In new dataset there were more images and slightly fewer tags (I combined tags that gave very similar results). Only then was this LoRA trained.<hr /><h3 id="disclaimer">Disclaimer</h3><hr />This model wasn't made for making pornography. The dataset included images of character that wasn't real. It also included images of naked fictional character, but those images were used only to improve the model's quality. I do not take responsibility for any images that was created or will be created using my model and then published by other people.

Files with SHA256: c2c0489d551e85ae52a6801925d9d09a4d722102121f8629d52b6c5cccdb5889