Regional Conditional Sampling — v5.0 DenseDiffusion

注意:

目前DenseDiffusion这个插件有BUG,某些出图尺寸会出现张量错误,解决方法是换成可用的尺寸,比如7041280 , 5122048等等,另外在进行放大时需要选择laten放大并选择合适的倍数进行放大,否则也会出现张量错误,经过测试发现,不同的出图尺寸需要不同的放大倍数才能正常进行放大,上面说过的2个尺寸,1.5倍和2倍没有问题,其他倍数请自行测试,二次放大时的倍数同样需要进行测试,比较尴尬,但是可行,至于像素放大目前我自己发现不可用,需要更多测试才能得出结论。

Attention:

The current DenseDiffusion plug-in has bugs, and tensor errors will occur in some drawing sizes. The solution is to replace it with available sizes, such as 704 1280, 512 2048, etc. In addition, when magnifying, you need to select laten to enlarge and choose the appropriate multiple to enlarge. Otherwise, tensor errors will also occur. After testing, it is found that different drawing sizes require different magnifications to enlarge normally. The above-mentioned 2 sizes, 1.5 times and 2 times are no problem. Please test other multiples yourself. The multiples at the time of secondary enlargement also need to be tested, which is awkward, but feasible. As for pixel enlargement, I have found that it is not available at present, and more tests are needed to reach a conclusion.

为了方便画面统一风格,增加字符串合并节点以优化了工作流

In order to facilitate the unified style of the screen, string merging nodes are added to optimize the workflow