IPAdapter Combine Weights V2

IPAdapter Combine Weights V2 is about as minimal as a node gets: it takes two sets of weights and glues them into one list. That's the entire job. It's a utility in the animation toolbox - when you're building a per-frame weight schedule by hand, sometimes you have "first half of the clip, ramping up" in one place and "second half, easing out" in another, and you need them as a single ordered list before the Batch apply nodes will accept them.

What it does

weights_1 and weights_2 are required FLOAT inputs (default 0.0, range 0 to 1 each). Combine each into a list - a bare float becomes a one-element list - then concatenates them in order. Outputs: weights, the merged FLOAT list, and count, the total length as an INT. That count is the part people actually use: it tells you how many frames your schedule now covers, so you can check it against your batch before you run and catch a length mismatch before it produces a hard cut instead of a fade.

When you'd use it

Realistically, you're only reaching for this if you're doing video or animation work and hand-building schedules, or if you're splitting a schedule across two IPAdapter Weights V2 nodes because one of them is doing something fancier per section (different easing for the intro and the outro). For a single continuous ramp you're better off using IPAdapter Weights V2 alone with a linear or ease timing - this node exists to assemble, not to generate. If you never animate, you can ignore it entirely.

The merged weights output is a plain FLOAT list, which plugs into the weights input of the Batch apply nodes (IPAdapter Batch (Adv.) V2, IPAdapter Embeds Batch V2, and the FaceID batch variant) - those apply weight i to frame i.

Installing it

ComfyUI Manager: search "ComfyUI_IPAdapter_plus_V2". Or:

cd ComfyUI/custom_nodes
git clone https://github.com/chflame163/ComfyUI_IPAdapter_plus_V2

Restart ComfyUI. No models, no dependencies beyond the pack itself.

Gotchas

There's nothing here to break - the only real trap is assuming the outputs are scaled back to 0–1 when they aren't: it's raw concatenation, so if list A tops out at 1.0 and list B starts at 0.3, there's an abrupt jump between them at the seam. If that matters, build the smooth transition into the source lists. And as with the rest of the pack, this is a renamed snapshot of cubiq's V2 codebase - a versioned copy, not an auto-updating fork.