<h2>ℹ️What is this | 这是什么</h2>这是一个风格LoCon / LoRA，可以画出具有繁复结构的类蒸汽朋克/赛博朋克/... 反正就是机械风格的角色，取决于你填什么提示词。风格来源于Dylan Kowalski。This is a style LoCon / LoRA which generate highly detailed, complex and steampunk-alike character. The artstyle is from 3D artist Dylan Kowalski.训练素材包含10张图片，分辨率512*768。训练基础模型为ChilloutDataset contains 10 pictures with resolution at 512*768, trained on chillout.预览图片生成使用<a target="_blank" rel="ugc" href="https://civarchive.com/models/21952/eris">Eris</a>，它和这个LoCon 配合的很好。Previews are generated with <a target="_blank" rel="ugc" href="https://civarchive.com/models/21952/eris">Eris</a>, which works very well for this LoCon.更新了一些图片示例 / Updated some images.添加了Epoch-06，对画风影响较弱。可自行尝试 / Added Epoch-06, these have lesser effect on style, for those who wish to reduce the artifacts.<h2>💦Known Issues | 已知问题</h2>🫱在部分Checkpoint下手画的非常答辩，建议在负面提示词中加入hand。Hands are really bad (at least on some checkpoints), avoid generating hands by adding "hand" into the negative prompts, or use "portrait, upper body, close up," in the positives.⚖️权重高于0.5会导致画面迅速崩坏，LoRA加载过多可能导致最终图片模糊。建议最高0.5，少加载LoRA。<a target="_blank" rel="ugc" href="https://github.com/klimaleksus/stable-diffusion-webui-anti-burn.git">这个插件</a>有可能缓解。Weight above 0.5 tends to destroy image. Loading too many LoRA will produce blurry images. <a target="_blank" rel="ugc" href="https://github.com/klimaleksus/stable-diffusion-webui-anti-burn.git">This </a>might help.👦有时候会拒绝画男性，在负面中添加“1girl, female”可以解决这个问题。Sometimes it refuses to generate male character, add "1girl, female" in the negative to solve this.<h2>❓How to use | 使用方法</h2>✅必须提示词 / Mandatory trigger：🔸byDylanKowalski,&lt;lora:dylanKowalski:0.5&gt;,✳️可以搭配以下提示词 / Works well with：🔸steampunk（for steampunk aesthetic）🔸cyberpunk（for cyberpunk aesthetic）🔸intricate details🔸ornaments🔸prosthesis🔸clockwork✳️其他和材质相关的提示词 / OTHER PROMTS INDICATE MATERIALS, BE CREATIVE. (EXAMPLE)🔸porcelain🔸titanium🔸gemstones🔸holographic🔸...<h2>📝其他信息与参数:</h2>🔸Model: Eris / Darelites Fantasy Mix🔸Image Resolution / 分辨率: 512*768🔸Sampler / 采样器: DPM++ 2M Karras,🔸Sampling Steps / 采样步数: 20-40,🔸Hires fix / 高分辨率修复算法: R-ESRGAN 4+.🧡 希望你喜欢这个模型。I hope you enjoy it as much as I do. 🧡

Files with SHA256: c9b140626537f53b8419eeb7c5839a5621acd2a345ef23f7170f8899d89acec3