Combine Conditioning

This is the configurable version of Sage_AverageConditioning. Both nodes take a list of already-encoded conditionings and merge them into one, but where Average has exactly one behavior baked in, this node exposes an operation dropdown that picks how the values get combined - the node's own tooltip describes it as selecting "how conditioning values should be scaled." If you need a specific blending behavior beyond a plain average - and you know what you're after - this is the node with the extra control; if you just want a straightforward blend, Average Conditioning does the same core job with one less decision to make.

Like Average Conditioning, this operates purely on tensors that already exist - it doesn't touch CLIP, doesn't take text, and doesn't care what produced the conditionings you feed it, as long as they came from a compatible model.

Inputs and outputs

• operation - a dropdown that selects the combination behavior. The schema doesn't spell out every option by name beyond "how conditioning values should be scaled," so the practical move is to try each choice on a real prompt and watch what changes rather than guess from the label alone.
• conditionings - an expandable list, same as Average Conditioning: add as many existing CONDITIONING outputs as you want combined.

Output is a single conditioning, the combined result under whatever operation you selected.

Installing it

Through ComfyUI Manager: search Sage Utils, install, restart. Manually:

cd ComfyUI/custom_nodes
git clone https://github.com/arcum42/ComfyUI_SageUtils
cd ComfyUI_SageUtils
pip install -r requirements.txt

Restart ComfyUI afterward. No extra downloads needed - this is a pure conditioning-math node.

Common issues

Same caveat as its sibling: only combine conditionings that came from the same CLIP model or architecture. Mixing conditionings from unrelated encoders is technically possible if the tensor shapes happen to line up, but the result is generally muddy rather than useful.

If you're not sure whether to reach for this node or Sage_CombineCLIPTextEncode instead, the distinction is where the combination happens: Sage_CombineCLIPTextEncode combines raw text before it ever touches CLIP, this node combines already-encoded conditionings after the fact. They can produce meaningfully different results on the same inputs - text-level combination lets the pieces interact inside CLIP's attention, conditioning-level combination keeps them independent until the final blend. If one approach isn't giving you what you want, the other is worth a try before concluding the idea itself doesn't work.