Combine Conditionings

Small utility, clear job. Combine Conditionings takes several CONDITIONING inputs and merges them into a single one. The README's description is the whole thing: "It takes multiple conditionings as input and combines them into a single conditioning." You use it when you've encoded a few prompts separately - maybe from different text encoders, maybe a base prompt plus a style prompt, maybe conditionings coming out of different branches - and you want to feed them to a sampler as one.

It's the kind of node you don't think about until you have three conditioning wires and one positive slot, and then it's exactly what you need.

How it works

ComfyUI's native "Conditioning (Combine)" node merges two conditionings; this is Impact Pack's take, built to gather several at once into a single output. Combining stacks the conditionings so the model is guided by all of them together during sampling - as opposed to concatenating, which joins them end-to-end into one longer sequence. That distinction matters, and the pack ships both: this node combines, while Concat Conditionings concatenates. If your merged prompt feels like the pieces are competing rather than blending, that's the combine-vs-concat choice talking.

The inputs and outputs

The schema shows one required input, conditioning1, but the node is built to take more - it's a variadic node, so additional conditioning slots (conditioning2, conditioning3, ...) appear as you connect them, the same pattern the pack's other multi-input nodes use. You wire in as many conditionings as you've got. The single output is CONDITIONING, ready to drop into the positive (or negative) slot of any sampler.

No numeric knobs, no modes. It merges what you give it.

How to install it

ComfyUI Manager: search ComfyUI Impact Pack, install, restart. Manual: cd ComfyUI/custom_nodes && git clone https://github.com/ltdrdata/ComfyUI-Impact-Pack, install requirements into ComfyUI's own Python (pip install -r requirements.txt, or ..\..\..\python_embeded\python.exe -m pip install -r requirements.txt on Windows portable), then restart. There's no auto-install since v7.6, so run the requirements. A small SAM model downloads to ComfyUI/models/sams on first launch - not needed by this node, just the pack's default setup.

Common issues

The result isn't what you expected - parts feel muddy. You may want concatenation, not combination. Combining runs the conditionings in parallel; concatenating joins their token sequences. Try Concat Conditionings if blended prompts are washing each other out, or vice versa.

Only one input connected. With a single conditioning wired in, this node just passes it through - harmless, but pointless. Connect at least two for it to do anything.

Encoder mismatch. Conditionings encoded by different CLIP models or different-architecture text encoders aren't guaranteed to combine sensibly. Keep the conditionings you're merging from a compatible encoding path unless you're deliberately experimenting.