Lora trained on Coven Ahri from League of LegendsUpdate: double the dataset, better base model (NAI), different kohya settings, trained on 768x768, lower steps better flexibilitySome pictures were done with ComfyUI since Civit finally supports its metadatatags: coven ahri, 1girl, dress, detached sleeves, cleavage, multicolored hair, blonde hair, black hair, animal ears, pantyhoseoptional tags: whisker markings, multiple tails, energy ballYou definitely need more tries on those ones since Stable Diffusion doesn't like countingnegative embeddings: <a target="_blank" rel="ugc" href="https://civarchive.com/models/71961/fast-negative-embedding-fastnegativev2">Fast Negative Embedding v2</a>Recommended settings:<ul><li>Make sure you are aware on the usage instructions of LORA</li></ul><ul><li>VAE: <a target="_blank" rel="ugc" href="https://huggingface.co/stabilityai/sd-vae-ft-mse-original/blob/main/vae-ft-mse-840000-ema-pruned.safetensors">vae-ft-mse-840000-ema-pruned</a> (for realistic models &amp; 2,5D), <a target="_blank" rel="ugc" href="https://huggingface.co/hakurei/waifu-diffusion-v1-4/blob/main/vae/kl-f8-anime2.ckpt">kl-f8-anime2</a>, <a target="_blank" rel="ugc" href="https://huggingface.co/NoCrypt/blessed_vae/blob/main/blessed2.vae.pt">blessed2.vae</a> (for anime models)</li></ul><ul><li>Sampler: DPM++ SDE Karras or DPM++ 2M Karras(euler a for img2img upscaling) from time to time im also using <a target="_blank" rel="ugc" href="https://civarchive.com/models/35966/dpm-2m-alt-karras-sampler">DPM++ 2M alt Karras</a></li><li>Steps: 20 to 35</li><li>Clipskip: 2</li><li>Upscaler : <a target="_blank" rel="ugc" href="https://github.com/pkuliyi2015/multidiffusion-upscaler-for-automatic1111">GitHub - pkuliyi2015/multidiffusion-upscaler-for-automatic1111</a> <a target="_blank" rel="ugc" href="https://mega.nz/folder/qZRBmaIY#nIG8KyWFcGNTuMX_XNbJ_g/file/vRYVhaDA">4x-UltraSharp(for 2,5D &amp; realistic)</a>, <a target="_blank" rel="ugc" href="https://huggingface.co/konohashinobi4/4xAnimesharp/tree/main">4xAnimesharp(for anime)</a></li><li>CFG Scale : 5 to 9 (mostly 7)</li><li>LORA weight for txt2img: 0.5-1 (for the examples i mostly used 0.8 and 1)</li></ul><ul><li>Denoise strength upscaling: 0.3 to 0.55</li></ul>If you want to autocomplete the tags &amp; trigger im using, I recommend using <a target="_blank" rel="ugc" href="https://github.com/DominikDoom/a1111-sd-webui-tagcomplete">Booru style tag autocompletion for AUTOMATIC1111's Stable Diffusion web UI</a>For full body shots i recommend using: <a target="_blank" rel="ugc" href="https://civarchive.com/models/33192/comfyui-impact-pack">ComfyUI Impact Pack</a>or the a1111 alternative <a target="_blank" rel="ugc" href="https://github.com/Bing-su/adetailer">GitHub - Bing-su/adetailer</a>

Files with SHA256: 11ddefa79da67cb4db23a8e46fe98beb94024234512effd8288b31782babd626