Weight Schedule Extend

If you've built animation-style workflows in ComfyUI - a LoRA strength that ramps up over a clip, a denoise value that changes per keyframe, anything driven by a value that varies across frames rather than staying fixed - you've probably run into the annoyance this node solves: two schedule nodes producing their output in two different shapes that don't want to talk to each other. One hands you a plain Python list, another hands you a pandas Series, a third wants a tensor. This node's job, per its own description, is to "extend, and convert if needed, different value lists/series" - joining two value series into one and normalizing the result to whatever format your next node actually expects.

How it works

You feed it two value series, input_values_1 and input_values_2, and it produces one combined series as output, converting types along the way if the two inputs didn't already match. Where it earns its keep is when you've built two separate weight curves - say one that handles the first half of a sequence and another for the second half - and the node downstream wants them as a single continuous series in a specific format, not two disconnected pieces. It's a joiner and a type-adapter in one node, which is a small thing, but it's exactly the kind of small thing that stops a schedule-heavy workflow from breaking on a type mismatch three nodes later.

Worth being straightforward about: beyond the node's own one-line description, there isn't much documentation on the exact extension behavior - whether it strictly concatenates the two series end-to-end or handles overlap differently isn't something this node's interface makes explicit. Treat it as a practical tool to reach for when you already have two schedules and a downstream type error, rather than something to build a workflow around from a standing start.

The inputs and output that matter

• input_values_1 and input_values_2 - your two value series. Both are typed as FLOAT and default to a bare 0, which tells you these inputs are meant to be fed by other schedule-producing nodes rather than typed in by hand - a schedule node upstream is doing the real work of building each curve.
• output_type - controls the format of the combined result: match_input keeps whatever type the first input arrived as, or you can force list, pandas series, or tensor regardless of what came in. Set this to whatever your next node in the chain actually expects.

The single output is the combined series, typed FLOAT the same way the inputs are - ComfyUI's type system doesn't have a dedicated "list of floats" pin, so schedule nodes across this pack pass array-like data through the plain FLOAT type by convention. Don't be surprised that a "float" output is actually carrying a whole series.

How to install it

Part of the KJNodes pack. Through ComfyUI Manager: search "KJNodes for ComfyUI," install, restart. Manually: cd ComfyUI/custom_nodes && git clone https://github.com/kijai/ComfyUI-KJNodes, then pip install -r ComfyUI-KJNodes/requirements.txt (portable Windows build: python_embeded\python.exe -m pip install -r ...), then restart. No extra models needed - this is pure data-shape utility work.

Common issues & troubleshooting

A downstream node rejects the output with a type error. Change output_type to whatever format that node expects - list, pandas series, or tensor - rather than leaving it on match_input and hoping it lines up.

You wired a plain number into input_values_1 or input_values_2 and got nothing useful. These inputs are meant to receive an actual series from a schedule-building node upstream, not a hand-typed constant. A bare 0 is just the placeholder default, not a realistic input.

You're not sure this node applies to your workflow. If you're not already juggling two separate per-frame or per-step value schedules that need joining, you probably don't need this node yet - it's a fix for a specific plumbing problem in schedule-heavy setups, not a general-purpose utility.