Nifty Random Noise
Hand a sampler its noise, instead of letting it make its own
- seed_actions
- NOISE
Samplers start from noise, and in ComfyUI the noise is usually generated for you, invisibly, from the seed you set on the KSampler. But some samplers - the advanced ones that accept a NOISE input - let you control that step directly. Nifty Random Noise is the node that generates the noise for you: give it a seed, get a NOISE object, hand it to a sampler that expects one.
What it is
Inputs:
- seed - the seed for noise generation (0 up to a huge max, default 0). Same seed, same noise, every time - that's the reproducibility contract.
- seed_actions - the pack's shared seed widget, with a 🎲 New Seed button that randomizes immediately and a Random toggle that makes the node roll a new seed every time you hit Queue (on by default).
Output: NOISE - a noise object ready for an advanced sampler node that takes a NOISE input.
How it works
Under the hood it wraps ComfyUI's own prepare_noise - the same generator the core KSampler uses internally. So you're not getting some exotic custom noise; you're getting the standard, battle-tested noise, but produced up-front and explicitly instead of implicitly. The point isn't a different noise, it's control: the noise is now a first-class value in your graph.
Why you'd reach for it
- Advanced samplers. Nodes like SamplerCustomAdvanced and other custom sampling nodes take a
NOISEinput, and they won't work without one. This is how you feed them. - Separating seed from sampler. When you want one seed value to drive noise across multiple sampler stages, generate the noise once here and split the
NOISEoutput to both - instead of hoping every KSampler reads the same seed. - Noise you can inspect. Because the noise is now a node output, you can preview or manipulate it if your workflow gets that deep.
It's the same seed math you already know from KSampler - the difference is only that the noise is explicit. That's genuinely useful for advanced sampling setups, and honestly invisible if you're only using the basic KSampler.
Installing it
Part of the Nifty Nodes pack:
cd ComfyUI/custom_nodes
git clone https://github.com/Stibo/comfyui-nifty-nodes
or search "Nifty Nodes" in ComfyUI Manager, then restart. No models, no dependencies beyond the pack.
Gotchas
The biggest surprise is the Random on Prompt behavior, which defaults to on: if you connect this node expecting a fixed seed, every queue run will silently roll a new seed. Toggle the Random button off if you want reproducibility. Second, this node is only useful with samplers that accept a NOISE input - wire it into a plain KSampler and it does nothing for you. And as with the whole pack, it targets ComfyUI's newer V3 API, so update ComfyUI if the node doesn't show up after install.
Inputs (2)
| Name | Type | Default | Description |
|---|---|---|---|
| seed | INT | 00–1125899906842624 | — |
| seed_actions | NIFTY_SEED_ACTIONS | — |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| NOISE | NOISE | — |