不要问为什么别的模型能出这个模型不能出，或者为什么细节/画面不丰富精美，你先看看你写没写或者写了什么奇怪的东西，没写的东西别指望模型给你出图。

————————————

Anything-Ink

hugging face:X779/Anything_ink · Hugging Face

1.原模型使用60w+张基于novelAI和nijijourney生成的图片在单张4090上训练约2个月，后续通过一部分“魔法”步骤得到现在所看到的ink模型

2.请避免使用例如只有“1girl”此类的提示词，模型提示词准确度极高，对使用者的prompt书写能力会有所要求。prompt可以参考：元素同典：确实不完全科学的魔导书

3.本模型不适配novelAI相关的LoRA模型，强行使用可能出古神“鬼图”

4.不建议使用任何外置VAE，否则容易出现图像失真或者“鬼图”

5.因为训练SDXL所需算力已经超过本人极限，且以目前情况来看可能即使训练提升也不会很大，所以目前不会考虑任何SDXL的训练计划。

*.这个模型单纯是因为我单独发出来没人看就标了个anything的名称放到这里了

1. The original model was trained on a single 4090 for about 2 months using more than 600,000 images generated based on novelAI and NijiJourney, and then the ink model seen now was obtained through a part of "magic" steps.

2.Please avoid using prompt words such as "1girl", the model prompt words are highly accurate, which will require the prompt writing ability of the user. The prompt writing can refer to the following document:元素同典：确实不完全科学的魔导书

3. This model is not suitable for LoRA model related to novelAI, and the forced use may result in "ghost pictures"

V5[PRT]

AnythingV5之后的模型并非只需要使用简单提示词就可以看似很好的图，也并非只有“1girl”，它需要精准的提示词以达到对应的效果。

The model after AnythingV5 doesn't just need to use a simple prompt to look good, and it doesn't just have "1girl"; it needs precise prompt words to achieve the corresponding effect.

————————————————————

模型训练工具︱Recommend a model training tool

7eu7d7/HCP-Diffusion-webui: webui for HCP-Diffusion