Icy ConditioningMultiply

IcyConditioningMultiply is the IcyHider clone of ConditioningMultiply, and it's the "prompt volume knob" of the conditioning toolbox. If your prompt feels like it's being ignored, or dangerously over-assertive, this node lets you scale its influence directly - with a range that can do more harm than good if you're not careful.

What it does

Two inputs, one output:

• conditioning - the conditioning you want to scale.
• multiplier - a float from -100 to 100, default 1.0. Multiply by 1.0 and it's a passthrough. By 2.0, the prompt's influence is doubled. By 0.5, halved.

Output is a CONDITIONING with the embedding values scaled by the multiplier. That's the whole mechanism: it doesn't re-encode anything, doesn't touch the sampler's other settings - it just makes one prompt pull harder or softer on the denoise.

The sane range is tiny

Here's the honest part, and it's the whole node in a sentence: the slider goes to 100, but the usable range is roughly 0.2–2.0, and negative values are a different feature entirely.

• 0.2–0.8 - softening a prompt that's over-locking the image, or dimming a background prompt so it's context, not demand.
• 1.0–2.0 - making a weak prompt more insistent. Beyond about 2, output starts to break down - oversaturated, glitchy, "the prompt is shouting" territory.
• 0 - effectively the same as an empty prompt; the conditioning stops steering.
• Negative - inverts the guidance: the model is pushed away from the prompt's content. This is the experimental corner. People use it for "generate anything but this," and sometimes it's brilliant, but it's the first thing to produce garbage in an otherwise-working graph.

Where people get burned: treating the slider as if 5× will make the prompt five times more obeyed. It won't - it'll wreck the sample. If your prompt is being ignored, the multiplier is rarely the fix; check token truncation, encoder compatibility, and negative-prompt inertness on distilled models before you reach for volume.

The "Icy" caveat

Outputs CONDITIONING - no image preview, so the Icy wrapper is cosmetic relabeling. Same core ConditioningMultiply code underneath; the pack just clones every core node.

Install

ComfyUI Manager → "IcyHider" → install → restart, or:

cd ComfyUI/custom_nodes
git clone https://github.com/icekiub-ai/ComfyUI-IcyHider.git

Restart. No Python deps, no models bundled. Node under IcyHider Comfy Core. Start at 1.0, move in small steps, and let the model tell you when you've found the edge.