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>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><h3>Trigger words for V1.20:</h3>positive: &lt;lora:Mermaidv220:1&gt;(mermaid, full body)You can also try "monster girl", "pussy", "green scales" (or any other color WITH LIMITED results), "head fins" LoHa is biased towards generating a "pussy" since the entire dataset had one! Use negative "pussy, nsfw" and "sfw" in positives for sfw gens!<h3>Follow the steps in the Model Version notes to reproduce my images.</h3>Sample images generated with Grapefruit.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 75 512x512 images and 20 epochs using Animefull-final-pruned for the Training Model.Danbooru tags generated by Waifu Diffusion 1.4 Tagger and manually fine tuned.

Mermaids - LoHa

Generate using a weight of 0.5-1.0 like this: &lt;lora:mermaidv215:0.75&gt;.Overall feels better with anatomy compared to first version, but still have problems with fins.Dataset reduced down. 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: aafd0674b1bd021a18bffa0ef4b1e0dba630bbae9c1ccd460c41d9444cea7820