Icy ConditioningZeroOut
The 'give the model nothing to work with' node — and when that's actually useful
- conditioning
- CONDITIONING
IcyConditioningZeroOut is the IcyHider clone of a node that barely gets talked about and is quietly important in the right graph: ConditioningZeroOut. As the name says, it takes a conditioning and replaces its contents with zeros - you feed it any CONDITIONING and you get back an empty conditioning, the text-prompt equivalent of a blank page.
The Icy prefix is pure packaging. IcyHider clones every core node at startup, this one included, and the only thing the wrapper changes is frontend behavior (which is irrelevant here - conditioning has no preview to hide). Use the core node and nothing about your workflow changes.
Why you'd want a zeroed conditioning
The one field is conditioning and the one output is a CONDITIONING. Boring on purpose. The use cases:
- An honest "no prompt" negative. When you want the unconditional pass to be actually unconditional - not "worst quality, blurry" steering, just nothing. If you're running a method that needs a real empty branch (some regional and attention-manipulation setups do), this is the honest way to build it rather than typing a random negative that secretly guides.
- Isolating a variable. Zeroing a conditioning is the quickest way to answer "how much of this image comes from this prompt?" - swap the branch in, regenerate on a fixed seed, and any change is attributable to the text.
- As a building block. Other conditioning utilities expect a conditioning input; when you need a neutral placeholder in a chain, ZeroOut supplies one without inventing content.
It's a scalpel, not a hammer. If your workflow doesn't need an empty conditioning, this node does nothing for you - and that's fine. Most graphs never touch it.
The honest "Icy" caveat
Because it outputs conditioning, the IcyHider hiding JavaScript has nothing to cover here. This wrapper exists only because the pack wraps all core nodes automatically - the author's nodes.py walks every class in ComfyUI's core module and subclass-relabels them. So if you're on the IcyHider install and this node just appeared in your menu alongside a dozen other Icy loaders, that's the mechanism doing its thing, not you having found something special.
Install
Get the pack once and every Icy node comes with it. ComfyUI Manager → search "IcyHider" → install, then restart. Or manually:
cd ComfyUI/custom_nodes
git clone https://github.com/icekiub-ai/ComfyUI-IcyHider.git
Restart ComfyUI. No Python dependencies (the pack's pyproject.toml ships empty) and no model downloads. Nodes land in the IcyHider Comfy Core category. If the whole Icy family is missing after install, you forgot the restart. And if you ever doubt what this node computes, don't - the wrapper is frontend-only; the actual zeroing is ComfyUI's own core code.
Inputs (1)
| Name | Type | Default | Description |
|---|---|---|---|
| conditioning | CONDITIONING | — |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| CONDITIONING | CONDITIONING | — |