Trigger token: sora \(arknights\)idol clothes for official outfit. Individual parts are not tagged, besides the cape, white thighhighs, boots. Gloves are baked when using the official outfit trigger, for now. tail or wolf tail are in the dataset and are reliable.microphone, microphone stand are in the dataset. I don't how reliable it is.Her <a target="_blank" rel="ugc" href="https://arknights.wiki.gg/wiki/Sora#/media/File:Sora_Skin_1.png">Summer Flowers</a> and <a target="_blank" rel="ugc" href="https://arknights.wiki.gg/wiki/Sora#/media/File:Sora_Skin_2.png">Melodiosa </a>outfit are in the dataset, both are currently unreliable. Both lazily tagged for now. (1*)LoRA name in preview images is different due to a typo; 6d41bc687ae2 is my local hash.Typical LoRA issues of some fine details being incorrect here and there, etc...Prompts and checkpoint models should be included in the examples images, with several different checkpoints used for compatibility.Generated with an upgraded/customized version of <a target="_blank" rel="ugc" href="https://github.com/Haoming02/sd-webui-forge-classic/tree/neo">Forge NEO</a> with Euler Normal. Upscaled by 1.5 with <a target="_blank" rel="ugc" href="https://openmodeldb.info/models/2x-AnimeSharpV3">2x-AnimeSharpV3</a> 0.55 denoising, Euler A Normal, + 0.5 to base CFG, and + 1 step from base.Trained on 47 images (1 AI Image) with <a target="_blank" rel="ugc" href="https://github.com/derrian-distro/LoRA_Easy_Training_Scripts">Derrian Distro</a>. Config is embedded into the model.(1*) I don't get paid to do this, these can take plenty of hours to make. My main focus was making another Sora LoRA, in default clothing, because most checkpoints I've used cannot.

Arknights Sora / ソラ(アークナイツ) SDXL

Full FP16, Net Dim 8, Net Alpha 6. Fourth model trained after testing.

SoraARK_v4Distro-000012.safetensors

v1.0

THIS IS A <a href="https://civitai.com/models/2376105" alt="Reprint url">REPRINT</a>!Go to original author to check their other models or to support them. <hr /> Trigger token: sora \(arknights\)idol clothes for official outfit. Individual parts are not tagged, besides the cape, white thighhighs, boots. Gloves are baked when using the official outfit trigger, for now. tail or wolf tail are in the dataset and are reliable.microphone, microphone stand are in the dataset. I don't how reliable it is.Her <a target="_blank" rel="ugc" href="https://arknights.wiki.gg/wiki/Sora#/media/File:Sora_Skin_1.png">Summer Flowers</a> and <a target="_blank" rel="ugc" href="https://arknights.wiki.gg/wiki/Sora#/media/File:Sora_Skin_2.png">Melodiosa </a>outfit are in the dataset, both are currently unreliable. Both lazily tagged for now. (1*)LoRA name in preview images is different due to a typo; 6d41bc687ae2 is my local hash.Typical LoRA issues of some fine details being incorrect here and there, etc...Prompts and checkpoint models should be included in the examples images, with several different checkpoints used for compatibility.Generated with an upgraded/customized version of <a target="_blank" rel="ugc" href="https://github.com/Haoming02/sd-webui-forge-classic/tree/neo">Forge NEO</a> with Euler Normal. Upscaled by 1.5 with <a target="_blank" rel="ugc" href="https://openmodeldb.info/models/2x-AnimeSharpV3">2x-AnimeSharpV3</a> 0.55 denoising, Euler A Normal, + 0.5 to base CFG, and + 1 step from base.Trained on 47 images (1 AI Image) with <a target="_blank" rel="ugc" href="https://github.com/derrian-distro/LoRA_Easy_Training_Scripts">Derrian Distro</a>. Config is embedded into the model.(1*) I don't get paid to do this, these can take plenty of hours to make. My main focus was making another Sora LoRA, in default clothing, because most checkpoints I've used cannot.

Files with SHA256: 951be0b0645aa663abcddd65f7db63836e52be5ff1774580407435fb73778f8a