--- license: cc-by-nc-sa-4.0 --- # AI 元火娘计划 ## v1 (Dreambooth) 使用 dreambooth 在 Novel AI final pruned 模型基础上训练,之后融合 10% 的 Anything 3.0 模型。 **在 A1111 的 Web UI 中使用时,需要在设置中打开 `Enable quantization in K samplers`,并设置 `Clip skip` 为 2。** ### 妹妹 预览图:![yanhuo_v1_dreambooth](./预览图/yanhuo_v1_dreambooth.png) ```text (yanhuo), 1girl, masterpiece, best quality, white hair, ahoge, snowy street, [smile], dynamic angle, full body, [blue eyes], flat chest, cinematic light Negative prompt: lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry Steps: 20, Sampler: Euler a, CFG scale: 10.5, Seed: 2416091869, Size: 1024x768, Model hash: 5e03450011, Clip skip: 2, ENSD: 31339 ``` 下载: - [ckpt](./yanhuo_v1_dreambooth_clip2_10k_fp16.ckpt) - [safetensors](./yanhuo_v1_dreambooth_clip2_10k_fp16.safetensors) 推荐起手式: ```text (yanhuo), 1girl, masterpiece, best quality, white hair, ahoge ``` ### 姐姐 预览图:![yanyuan_v1_dreambooth](./预览图/yanyuan_v1_dreambooth.png) ```text (yanyuan), 1girl, (masterpiece), best quality, beautiful detailed sky, (park), sunset, [smile], dynamic angle, full body, flat chest, volume light, red eyes Negative prompt: lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry Steps: 20, Sampler: Euler a, CFG scale: 10.5, Seed: 2355417689, Size: 1024x768, Model hash: 270337850e, Clip skip: 2, ENSD: 31339 ``` 下载: - [ckpt](./yanyuan_v1_dreambooth_clip2_5k_fp16.ckpt) - [safetensors](yanyuan_v1_dreambooth_clip2_5k_fp16.safetensors) 推荐起手式: ```text (yanyuan), 1girl, masterpiece, best quality ``` ### 姐妹同屏 预览图:![yuanhuo_v1_dreambooth](./预览图/yuanhuo_v1_dreambooth.png) ```text (yanhuo with white hair and blue eyes and ahoge), (yanyuan with black hair and red eyes), 2girls, masterpiece, best quality, sisters, dynamic angle, flat chest, [smile], (steampunk: 1.5), Negative prompt: lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry Steps: 20, Sampler: Euler a, CFG scale: 7, Seed: 304955900, Size: 1024x768, Model hash: 15449d01d2, Clip skip: 2, ENSD: 31339 ``` 下载: - [ckpt](yuanhuo_v1_dreambooth_clip2_5k_fp16.ckpt) - [safetensors](./yuanhuo_v1_dreambooth_clip2_5k_fp16.safetensors) 推荐起手式: ```text (yanhuo with white hair and blue eyes and ahoge), (yanyuan with black hair and red eyes), 2girls, masterpiece, best quality, sisters ``` 姐妹同屏模型出图成功率比较低,容易出现瞳色混淆、衣服互换等情况,一般需要 img2img + 分区 inpaint 来反复修正。 ## 更新日志 - 2023.01.30:第一版模型。 ## 协议 本仓库所包含模型以署名-非商业性使用-相同方式共享 4.0 国际 ([CC BY-NC-SA 4.0]([Creative Commons — 署名-非商业性使用-相同方式共享 4.0 国际 — CC BY-NC-SA 4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/deed.zh))) 协议公开发布。 模型训练所用数据集来源于元火动漫社社员的创作,版权归原作者与元火动漫社所有,不予公开。