【v1.1 】重新整理素材，删掉了一些质量低的图片，然后再训练了一遍，效果应该比1.0更好。<hr />更新了一个用 Lokr 训练的版本，不一定比之前的更好：An updated version trained with Lokr is not necessarily better than the previous one:<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/999a66f1-f4ef-40be-b3e1-046c3303334a/width=525/999a66f1-f4ef-40be-b3e1-046c3303334a.jpeg" /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/050d9c05-8943-4327-b18f-26780df4d6c6/width=525/050d9c05-8943-4327-b18f-26780df4d6c6.jpeg" /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/79dda4a0-da91-457f-a31e-ecf5bb3ec196/width=525/79dda4a0-da91-457f-a31e-ecf5bb3ec196.jpeg" /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/133535a9-9aea-41bc-a373-6c9330a05ec4/width=525/133535a9-9aea-41bc-a373-6c9330a05ec4.jpeg" /><hr />这个模型是从 <a target="_blank" rel="ugc" href="https://civarchive.com/models/28169?modelVersionId=154252">https://civarchive.com/models/28169?modelVersionId=154252</a> 训练 Loha 得来的，我把它以 0.65 的 weight 合并为一个 checkpoint，所以需要额外的 VAE，推荐：<a target="_blank" rel="ugc" href="https://civarchive.com/models/22354/clearvae。">https://civarchive.com/models/22354/clearvae。</a>没有什么特别推荐的参数，可以自己尝试，hires fix 的效果很不错。This model is obtained from training Loha at <a target="_blank" rel="ugc" href="https://civarchive.com/models/28169?modelVersionId=154252">https://civarchive.com/models/28169?modelVersionId=154252</a>. I merged it into a checkpoint with a weight of 0.65, so additional VAE is required. Recommended: <a target="_blank" rel="ugc" href="https://civarchive.com/models/22354/clearvae">https://civarchive.com/models/22354/clearvae</a>.There are no particularly recommended parameters, you can try it yourself, hires fix works very well.

Files with SHA256: 99bdcb299e29212b018243b1bddf9c0d56141ba10f592f9921e5a44eae1ef181