Model used: <a target="_blank" rel="ugc" href="https://civarchive.com/models/9942/abyssorangemix3-aom3">AbyssOrangeMix3</a>Recommended VAE: orangemix or kl-f8Recommended LoRA weight: 0.8-1Trigger words: king, anime, one punchRecommended words in prompt: 1boy, blonde hair, solo, scar, blue eyes, long hairNegative prompt example:<pre><code>(worst quality, low quality:1.4), 3d, greyscale, monochrome, text</code></pre>On all images from preview I used DPM++ 2M SDE sampler, but you can use what you like.All other settings the same as for AOM model:Steps:<ul><li>DPM++ SDE Karras: Test: 12～ ,illustration: 20～</li><li>DPM++ 2M Karras: Test: 20～ ,illustration: 28～</li></ul>Clipskip: 1 or 2Upscaler :<ul><li>Detailed illust → Latenet (nearest-exact) Denoise strength: 0.5 (0.5~0.6)</li><li>Simple upscale: Swin IR, ESRGAN, Remacri etc… Denoise strength: Can be set low. (0.35~0.6)</li></ul>This LoRA was trained on images which I scraped from <a target="_blank" rel="ugc" href="http://fancaps.net">fancaps.net</a> (I made tampermonkey <a target="_blank" rel="ugc" href="https://github.com/Maximax67/Fancaps-Tampermonkey-Image-Downloader">extension</a> which adds a download button under each image). I used <a target="_blank" rel="ugc" href="https://github.com/toriato/stable-diffusion-webui-wd14-tagger">wd14 tagger</a> extension for Stable diffusion to automatically tag my images and <a target="_blank" rel="ugc" href="https://colab.research.google.com/github/hollowstrawberry/kohya-colab/blob/main/Lora_Trainer.ipynb">LoRA trainer colab</a> for training (42 images with 6 repeats, 2 batch size, 10 epochs).I plan to do a lot of LoRAs based on the One Punch Man anime. If you are interested, you can follow me.

King (One punch man)

Initial release of LoRA trained on 42 images from <a target="_blank" rel="ugc" href="http://fancaps.net">fancaps.net</a>.

Files with SHA256: 4e2fa4c3d5738eaf4e79848d0933cb01a26b73b5bc01f2cc4ae253da357fc5ed