Please check out my <a target="_blank" rel="ugc" href="https://ko-fi.com/part_loras">Ko-Fi</a> page if you are interested in supporting me.LoRA commissions open! -&gt; <a target="_blank" rel="ugc" href="https://www.fiverr.com/partai/develop-your-custom-lora-or-loha-model-for-stable-diffusion">Fiverr</a><h3 id="heading-85">This is a LoHa LoRA! You need the <a target="_blank" rel="ugc" href="https://github.com/KohakuBlueleaf/a1111-sd-webui-locon">extension</a>! Why? They do a lot better with multi-concepts.</h3>This is retrained from <a target="_blank" rel="ugc" href="https://civarchive.com/user/thesauceer">thesauceer</a> dataset, just a version of his <a target="_blank" rel="ugc" href="https://civarchive.com/models/19235/tutuballet-outfit">Tutu LoRA</a>, please check out his original work! He was kind enough to share his dataset.<h3 id="heading-86">Trigger words for V3.20:</h3>positive: &lt;lora:Tutuv320:0.5&gt;(tutu)You can also try WITH VERSION V2.20: "ribbon-trimmed skirt", "layered skirt" and "miniskirt". If failing to make full shirt, try adding either "leotard" or "tank top" or "covered navel" to prompt.Keep in mind you might have to adjust LoRA and/or tag weight for it to appear, defaults to "ribbon-trimmed skirt" most of the time! Dataset wasn't an exactly split between the different concepts.<h3 id="heading-87">Follow the steps in the Model Version notes to reproduce my images.</h3>Sample images generated with Hassaku.All negative embeds should be easy to find. and Prompts should be included in the sample images! Click the little icon in the bottom right to see it.It was trained on 45 512x512 images provided by thesauceer, 20 epochs using Animefull-final-pruned for the Training Model.Danbooru tags generated by Waifu Diffusion 1.4 Tagger.

Tutu/Ballet outfit - LoHa

Generate using a weight of 0.3-1.0 like this: &lt;lora:Tutuv220:1&gt;.Trimmed and improved dataset and new training settings on an attempt to improve consistency and quality of ribbon skirts. How I archived my generations:<ol><li>txt2img: DPM++ SDE Karras with 20 steps, 512x512(can go up to 768 for wide or tall), Restore faces on (using CodeFormer weight 1.0 in Settings), Hires-fix on, to 1024x1024 (double original size) with your preferred upscaler and Denoise 0.4, CFG 7. Then send to img2img!</li><li>img2img: DPM++ SED Karras with 20 steps, Restore faces on (CodeFormer weight 1), rescale to 1536x1536, CFG 7, Denoise between 0.3-0.7 (based on how much you want to improve the image), SAME PROMPT! Send to extras when happy with img2img/inpaint results.</li><li>extras: Scale to x3 (final will be 4608x4608, I suggest going lower, I did x3 just for the samples here) Upscaler 1 - FatalAnime 4x, Upscaler 2 - SwinIR 4x with 0.10 visibility, GFPGAN 0.10 visibility, CodeFormer 0.10 visibility.</li></ol>If anyone is curious about my version names: v{version}.{trainingEpochs}

Files with SHA256: 6faf06ad78464212351c2d3399ad0c1f034445ad9fdcb920663e86dc5fb8c5d2