REMADE Batch Image Blend

The straightforward one in the batch. REMADE Batch Image Blend takes two batches of images, images_a and images_b, and blends each matching pair together at a single blend_percentage. Frame one of A with frame one of B, frame two with frame two, and so on, until the whole batch is a new batch of crossfades. It's the "average two versions of the same thing" node, applied at batch scale.

Where you'd use it: you generated two variants of a shot and want a smooth middle ground; you want to fade a result toward a reference at a fixed strength across a whole set; you're interpolating between an original and an edited pass. The core ComfyUI blend nodes are single-image affairs, so if you're doing it for fifty frames, this saves you fifty copies of the same node.

How it works

Mechanically it's a solid-color mask trick. For each pair it builds a solid gray mask at round(blend_percentage * 255), inverts it, and runs Image.composite(a, b, mask). Since the mask value ends up dark (255 − percentage), the result is weighted toward b... which brings us to the counterintuitive bit:

blend_percentage behaves backwards from how you'll read it. The default is 0.8, and 0.8 means the result is ~80% images_b and ~20% images_a. The dial reads "how much A to keep" in most blend nodes; here it effectively reads "how much B." Crank it to 1.0 and you get almost all B. Set it to 0.2 for "mostly A." It's not a bug - it's just authored upside-down from intuition, and it will absolutely make you second-guess your own workflow the first time.

Inputs and outputs

• images_a (IMAGE) - first batch.
• images_b (IMAGE) - second batch. Batch lengths must match (asserted).
• blend_percentage (FLOAT, 0.0–1.0, default 0.8) - the blend weight, with the backwards semantics above.
• Output: images (IMAGE) - one blended frame per pair.

The constraints

• Same dimensions required. Unlike some nodes in this pack, there's no resize anywhere in this one - both images in a pair must already be the same size, or Image.composite throws. Match your resolutions upstream.
• Batch lengths must match, or the assert stops you at the first pair.
• It's a flat crossfade, not an alpha-aware composite. If the inputs carry transparency, composite treats it as pixels, which is usually not what you want - flatten first if you care.

Installing it

One of 12 nodes in Remade_nodes:

cd ComfyUI/custom_nodes
git clone https://github.com/Pheat-AI/Remade_nodes

Restart ComfyUI, or search Remade_nodes in ComfyUI Manager. No models; declared dependency is numpy (OpenCV must be present for the color-blend node or the pack won't load - pip install opencv-python if it does).

A perfectly serviceable batch crossfader once you accept the parameter inversion. Remember: 0.8 is mostly B, match your sizes, and you've got a fifty-frame interpolation in one node.