Feature Request: Mask painting precision assistance

还没有人认领这个 Issue。

Problem:

Currently, when mask is set to true, we have a blank canvas added to the right side of the input (or preview) image.

We then paint the areas we wish to mask, these masked areas are applied to the input image using the selected controlnet.

One issue with this is that it can be quite difficult to work out whether we are precisely masking the areas of the input image as intended, we have to kind of roughly judge, and if the desired masking area is small or needs to be precise this can be quite difficult.

Proposed solution:

It would be very helpful if one of the following could be implemented:

Either

a) Drawing a mask area in the mask image window also draws the painted mask area, @ low/medium opacity over the input image to the left.

or (preferable)

b) The input image can be shown/overlaid (with a toggle) in the mask image window, @ low/medium opacity, and be painted over, without this overlaid image interfering with or influencing the generated mask.

首先定位输入图像或预览图像以及蒙版图像窗口的 UI 代码，然后跟踪绘制的蒙版区域如何通过所选的 ControlNet 应用。比较提出的两种叠加方式，并定义一个可切换的低或中等不透明度输入叠加层，该叠加层不会改变生成的蒙版；完成的标准是能够看到精确对齐，同时蒙版处理的行为保持不变。

由索引模型根据 Issue 内容生成。