Stable Diffusion

Relolahricoven

model_id: relolahricoven

<p>Hey everyone!<br>I'm excited to showcase this fantastic SD model, created by the incredibly talented <strong><a href="https://civitai.com/user/AhriMain/models">AhriMain</a></strong>. All credit for this model goes to them. 🎨</p> <p>It's essential to support original creators, and if you're as impressed with their work as I am, check out their official page and show them some love: <strong><a href="https://civitai.com/user/AhriMain/models">AhriMain</a></strong>.</p> <p>Transparency and respect for artists are vital - If <strong><a href="https://civitai.com/user/AhriMain/models">AhriMain</a></strong> or anyone related feels uncomfortable with this model being showcased here, please reach out, and we'll take the necessary steps.</p> <p>---------------------------------------------</p> <p><strong>Original Model Description</strong></p> <p>&nbsp;</p> <div class="mantine-150s3z1 mantine-Spoiler-content" style="max-height: 962.8px;"> <div> <div class="mantine-TypographyStylesProvider-root mantine-1c4hq9n"> <div> <p>Lora trained on Coven Ahri from League of Legends</p> <p>&nbsp;</p> <p>Update: double the dataset, better base model (NAI), different kohya settings, trained on 768x768, lower steps better flexibility</p> <p>&nbsp;</p> <p>Some pictures were done with ComfyUI since Civit finally supports its metadata</p> <p>&nbsp;</p> <p>tags: coven ahri, 1girl, dress, detached sleeves, cleavage, multicolored hair, blonde hair, black hair, animal ears, pantyhose</p> <p>&nbsp;</p> <p>optional tags: whisker markings, multiple tails, energy ball</p> <p>You definitely need more tries on those ones since Stable Diffusion doesn't like counting</p> <p>&nbsp;</p> <p>negative embeddings: <a href="https://civitai.com/models/71961/fast-negative-embedding-fastnegativev2" target="_blank" rel="noopener ugc">Fast Negative Embedding v2</a></p> <p>&nbsp;</p> <p>Recommended settings:</p> <ul> <li> <p>Make sure you are aware on the usage instructions of LORA</p> </li> </ul> <ul> <li> <p>VAE: <a href="https://huggingface.co/stabilityai/sd-vae-ft-mse-original/blob/main/vae-ft-mse-840000-ema-pruned.safetensors" target="_blank" rel="noopener ugc">vae-ft-mse-840000-ema-pruned</a> (for realistic models &amp; 2,5D), <a href="https://huggingface.co/hakurei/waifu-diffusion-v1-4/blob/main/vae/kl-f8-anime2.ckpt" target="_blank" rel="noopener ugc">kl-f8-anime2</a>, <a href="https://huggingface.co/NoCrypt/blessed_vae/blob/main/blessed2.vae.pt" target="_blank" rel="noopener ugc">blessed2.vae</a> (for anime models)</p> </li> </ul> <ul> <li> <p>Sampler: DPM++ SDE Karras or DPM++ 2M Karras(euler a for img2img upscaling) from time to time im also using <a href="https://civitai.com/models/35966/dpm-2m-alt-karras-sampler" target="_blank" rel="noopener ugc">DPM++ 2M alt Karras</a></p> </li> <li> <p>Steps: 20 to 35</p> </li> <li> <p>Clipskip: 2</p> </li> <li> <p>Upscaler : <a href="https://github.com/pkuliyi2015/multidiffusion-upscaler-for-automatic1111" target="_blank" rel="noopener ugc">GitHub - pkuliyi2015/multidiffusion-upscaler-for-automatic1111</a> <a href="https://mega.nz/folder/qZRBmaIY#nIG8KyWFcGNTuMX_XNbJ_g/file/vRYVhaDA" target="_blank" rel="noopener ugc">4x-UltraSharp(for 2,5D &amp; realistic)</a>, <a href="https://huggingface.co/konohashinobi4/4xAnimesharp/tree/main" target="_blank" rel="noopener ugc">4xAnimesharp(for anime)</a></p> </li> <li> <p>CFG Scale : 5 to 9 (mostly 7)</p> </li> <li> <p>LORA weight for txt2img: 0.5-1 (for the examples i mostly used 0.8 and 1)</p> </li> </ul> <ul> <li> <p>Denoise strength upscaling: 0.3 to 0.55</p> </li> </ul> <p>&nbsp;</p> <p>If you want to autocomplete the tags &amp; trigger im using, I recommend using <a href="https://github.com/DominikDoom/a1111-sd-webui-tagcomplete" target="_blank" rel="noopener ugc">Booru style tag autocompletion for AUTOMATIC1111's Stable Diffusion web UI</a></p> <p>For full body shots i recommend using: <a href="https://civitai.com/models/33192/comfyui-impact-pack" target="_blank" rel="noopener ugc">ComfyUI Impact Pack</a></p> <p>or the a1111 alternative <a href="https://github.com/Bing-su/adetailer" target="_blank" rel="noopener ugc">GitHub - Bing-su/adetailer</a></p> </div> </div> </div> </div>