Conditioning Multiplier PoP

Conditioning is the tensor your text encoder produces and your sampler reads to know what you asked for - and once you're doing anything past a single positive prompt (combining conditionings, blending regions, weighting one text branch against another), you sometimes want to turn the volume up or down on one of those branches without touching the prompt text itself. That's what this node does: it takes a conditioning and multiplies it by a number. Nothing fancier than that, and the author says as much in the README - this isn't solving a deep problem, it's a volume knob for an existing conditioning signal.

How it works

It scales both parts of the conditioning tensor: the main embedding and the pooled output, when present, both get multiplied by multiplier. At 1.0 (the default) it's a no-op - conditioning passes through unchanged. Above 1.0, that branch's influence on the sample gets amplified relative to whatever else you're combining it with; the author specifically mentions using it "for adjusting the conditioning values with more complicated workflows" - think ConditioningCombine or ConditioningConcat setups where you want one input to dominate or recede rather than contribute equally.

Negative values are technically allowed, and the README is upfront that this is more a curiosity than a recommendation: "I wouldn't suggest doing this and haven't found any use for doing such a thing, but you have the option." Stick to positive multipliers unless you're specifically experimenting.

The inputs and outputs that matter

• conditioning - the conditioning you want to scale, from any CLIP Text Encode or upstream conditioning node.
• multiplier (default 1, range −1 to 3) - the actual dial. Above 1 amplifies, below 1 dampens, 1 does nothing.
• Output is a single CONDITIONING - drop it back into whatever expects conditioning: your KSampler's positive/negative slot, a ConditioningCombine node, an area-conditioning setup, wherever.

How to install it

ComfyUI Manager: search "comfy_PoP", install, restart. Manual:

cd ComfyUI/custom_nodes
git clone https://github.com/picturesonpictures/comfy_PoP

No dependencies to speak of - this is a pure tensor operation on data ComfyUI already produces, nothing to download.

Common issues & troubleshooting

Nothing changed after adding the node. Check multiplier isn't sitting at the default 1 - that value is a genuine pass-through, by design, so if you dropped the node into your graph and left it alone, you should expect no visible difference.

Output degrades or goes incoherent at high multiplier values. This isn't a bug - you're literally scaling up the magnitude of the conditioning tensor beyond what the model was trained to expect, and diffusion models get unstable fast outside their normal input range. Move in small steps (1.2, 1.4...) rather than jumping straight to 3.

Tried a negative multiplier out of curiosity. So did the author - per the README, don't expect it to do anything useful. It's exposed because the range allows it, not because there's a known good use for it.

Not sure if you want the multiplier or the normalizer. If your problem is "I want this conditioning branch to matter more or less," use this node. If your problem is "the pooled output feels too intense or inconsistent across generations," that's ConditioningNormalizer_PoP instead - a different tool for a different symptom.