PS Transfer

This is the node in Cyber-BlackCat's MoneyMaker pack that actually earns the "money" in the name. "PS Transfer" gives you eleven classic blend modes - add, multiply, screen, overlay, soft_light, hard_light, color_dodge, color_burn, difference, exclusion, divide - applied between two images, with an optional mask and a blur that does more than it looks like it does. If you only install one node from this pack, it's this one.

How it works

The node takes your source and target, Gaussian-blurs both with a sigma you control, and then applies the selected mode's math. That blur is the twist. The README's claim about "preserve artwork details such as text" comes from the add mode, which is really (source − blur(source)) + blur(target) - it extracts the high-frequency detail from the source (an unsharp-style residual) and lays it onto the target. That's how you push crisp text or linework from one image onto another without dragging along the underlying shading.

The rest are your standard compositing workhorses: multiply and hard_light for shading passes, overlay for that classic contrast bump, screen for lightening, difference for inverted blends. color_dodge and divide push highlights hard and clip. The whole thing is clamped to 0–1 at the end.

The inputs that matter

• target - the base image you're modifying.
• source - what you're layering on top (the detail, texture, or light pass).
• mode - the dropdown with the eleven modes; add is the default and the one with the detail trick.
• blur_sigma - Gaussian sigma for the pre-blur, default 1.0. This is the creative dial.
• blend_factor - how much of the blend to apply, from 0 (untouched target) to 1 (full blend), with a −10…10 range so negative values invert the effect.
• mask (optional) - pass a mask and the blend only applies where the mask is bright, turning the whole thing into a localized relight or grade. This is what makes it useful for compositing.

Output is a single IMAGE, ready to feed straight back into your graph.

Where people get burned

color_dodge and divide blow out. The result gets clamped to 0–1, so highlights clip hard and smooth gradients can band. Sometimes that's exactly the look - just know it's coming before you wonder why your sky is posterized. blur_sigma drives the Gaussian kernel size (6σ + 1), so a big value means a big, slow blur and mush; stay near 1–3 for detail work. And the source gets resized to match the target if the dimensions differ, which is convenient until it silently degrades a high-res source into a low-res target - match sizes upstream if that matters to you.

Honest take

There's overlap with ComfyUI's built-in blend utilities, but having all eleven modes in one node, with the blur and the mask, is convenient enough that I keep reaching for it. The add detail-transfer trick is the standout - genuinely useful for relighting and compositing work where you want someone's linework or text to survive the transfer. Slightly rough around the edges (the console will tell you the source was resized, and clipping is on you), but it does real work.

Install

Standard for the MoneyMaker pack. ComfyUI Manager - search "MoneyMaker" (pack title "ComfyUI-Yuan") - or by hand:

cd ComfyUI/custom_nodes
git clone https://github.com/Cyber-BlackCat/ComfyUI-MoneyMaker

Then restart ComfyUI. Dependencies are torch, torchvision, numpy, pillow, opencv-python, and scikit-image; scikit-image is the one most likely to be missing from an existing install, and Manager or pip install -r requirements.txt in your ComfyUI environment sorts it out. No model files anywhere in the pack.