图像混合模式

If you've ever wanted to layer a texture over a render and have it behave like a Photoshop blend layer - multiply it in, screen it in, hue-shift it - this is the node. MuyeImageBlendingMode takes a 前景图像 (foreground), a 背景图像 (background), and composites them with any of 27 classic blend modes: the whole normal/darken/lighten family (正片叠底 / multiply, 滤色 / screen, 叠加 / overlay, 柔光 / soft light...), the dodge-and-burn set (颜色加深, 线性减淡...), the difference/exclusion/sum/divide row, and the hue/saturation/color/luminosity group. It's a deterministic pixel operation, and the KB's post-processing doc makes the case that this is exactly what you should reach for instead of a diffusion pass - a blend mode is milliseconds; an img2img pass rewrites your face.

The controls are the standard blend stack, plus a couple of thoughtful extras:

• 系数 - overall blend strength, 0 to 1. At 1.0 the mode applies at full strength; lower it and the result drifts toward the background. This is your "how much of the texture do I actually want" knob.
• 数量差异补齐 - what happens when the two images have different batch sizes. 无 errors/requires matching, 循环 cycles the smaller batch, 首张补齐 / 尾张补齐 pad with the first or last frame, and 随机补齐 fills with random frames. If you're blending a texture across a whole image batch, 循环 is usually what you want.
• 色彩空间 - Lab or HSV. Matters mainly for the hue/saturation/color modes, where the blend happens in color space rather than RGB. Lab tends to give cleaner hue-shifts on skin; HSV is more aggressive. Both are real options, not cosmetic.
• 输出格式 - RGB or RGBA, for when you need the alpha through.

Output is one IMAGE, 混合图像. It auto-resizes the background to the foreground's size if they differ, and preserves alpha where present.

It's a genuinely capable node, and it pairs well with overlays - the pack's own TextOverlayImage produces text-on-image output that you can then blend over a background with this node. Where you'll feel the dependency: it uses scikit-image for the HSV conversion, so that one's a real install requirement.

Install

ComfyUI Manager → search "muye" → install "ComfyUI-Muye-nodes", restart. Manually:

cd ComfyUI/custom_nodes
git clone https://github.com/muyexiuluo/ComfyUI-Muye-nodes
cd ComfyUI-Muye-nodes
pip install -r requirements.txt

then restart. If the node is missing, pip install scikit-image is the usual fix. The README's manual block references an older ComfyUI_Muye.git URL; use the -nodes one above. Labels are Chinese by design (display name 图像混合模式; 混合模式 is the mode dropdown, 系数 is the strength). Missing node → check the console for [Muye] Failed to load module.