If you enjoyed this model, why don't you check out my other works? I put a lot of effort into correctly recreating artist styles and making the occasional character LoRA.😊You can also buy me a Coffee ☕<a target="_blank" rel="ugc" href="https://emojipedia.org/hot-beverage"> </a>at <a target="_blank" rel="ugc" href="https://ko-fi.com/stickyricky">https://ko-fi.com/stickyricky</a>Update:This update adds the definite version for most checkpoints but works best with the basic anime checkpoint: <a target="_blank" rel="ugc" href="https://huggingface.co/a1079602570/animefull-final-pruned/blob/main/model.ckpt">model.ckpt · a1079602570/animefull-final-pruned at main (</a><a target="_blank" rel="ugc" href="http://huggingface.co">huggingface.co</a><a target="_blank" rel="ugc" href="https://huggingface.co/a1079602570/animefull-final-pruned/blob/main/model.ckpt">)</a>This model was trained for NAI/AOM3 using the collab/guide by holostrawberry <a target="_blank" rel="ugc" href="https://civarchive.com/models/22530/guide-make-your-own-loras-easy-and-free">[Guide] Make your own Loras, easy and free - colabs | Stable Diffusion Other | Civitai</a>There already exist other LoRAs for this style on civitai and huggingface, but I always had some problems with them being oberbaked, washed out, resulting in bad anatomy or needing a huge number of sampling steps, when used with my preferred checkpoint AOM3, so I made my own.<h3 id="heading-1473">About the artist:</h3>Pixiv Fanbox: <a target="_blank" rel="ugc" href="https://msmspc.fanbox.cc/">MだSたろう｜pixivFANBOX</a>Twitter: <a target="_blank" rel="ugc" href="https://twitter.com/MsMspc">MだSたろう (@MsMspc) / Twitter</a>Mda Starou is a hentai artist, so you can expect the occassional nudity even when not specifically prompted. Here are some example images from the dataset the partial inspirations for two of the example images:<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/ea17400a-e30e-498d-be73-d3d445afaf32/width=525/ea17400a-e30e-498d-be73-d3d445afaf32.jpeg" /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/79202d03-bd04-4681-9c7d-ca2ccbc3aa08/width=525/79202d03-bd04-4681-9c7d-ca2ccbc3aa08.jpeg" /><h3 id="heading-1474">Weights &amp; Prompts</h3>The only necessary prompt is<pre><code>mda_starou</code></pre>v2.0 LoCon:checkpoint used in the examples:<a target="_blank" rel="ugc" href="https://huggingface.co/a1079602570/animefull-final-pruned/blob/main/model.ckpt">model.ckpt · a1079602570/animefull-final-pruned at main (</a><a target="_blank" rel="ugc" href="http://huggingface.co">huggingface.co</a><a target="_blank" rel="ugc" href="https://huggingface.co/a1079602570/animefull-final-pruned/blob/main/model.ckpt">)</a>Use a weight of 1.0. Adjust to taste and depending on your checkpoint (0.7-1.5).Character LoRA should generally be used at their lowest viable weight.This version is a huge upgrade compared to previous ones and should work with most anime style checkpoints.Explicit content in particular works with many prompts without adding position LoRA or similar. Be aware that the height to width ratio and sampling steps heavily affect the results.Concept LoRA will often introduce artifacts and should be your last resort.LoRA:Use a weight of 1.0. Go a bit lower (I used 0.75) for character loras if clothing gets fragmented or masks start showing up.LoCon:Use a weight of 1.0, adjust to taste. This model is more flexible and is compatible with others at full weight. It also hits their interpretation of FGO characters spot on.LoCon (NAI):Use a weight of 1.0, adjust to taste.For inspiration check the examples. I heavily recommend experimenting with the steps as well. Sometimes a 12 step image is just what you want, sometimes it takes 30 to get the necessary detail.For backgrounds it is rather lightweight, nudging images stylistically more towards anime.Disclaimer: This artist is tagged as creating Loli art. Creating and or sharing nsfw content of underage girls is not endorsed in any way.Please leave a review if you enjoyed this model or have problems with it. I'd also be interested in how the model behaves with other base checkpoints, so why don't you give it a try and throw some images down below! 👇

Mda Starou (MだSたろう) - Artist Style (LoRA/Lycoris)

155 images trained with 3 repetitions per generation in batches of 3, learning rate of 3e-4This version is a LoCon, not a LoRA, and needs an extension to work!

Files with SHA256: e34a55826fc4e7d785993c7f94774677f6313d0f77a489ff2952088d361b01d5