XY Input: Add/Return Noise
Add/Return Noise
- X or Y
A niche but specific tool: this node plots the two boolean noise-handling toggles from the advanced KSampler - add_noise and return_with_leftover_noise - so you can see, side by side, what each on/off state does to your output. It only makes sense in a multi-stage sampling setup, which is exactly where those toggles live. It's part of the Efficiency Nodes XY Plot system.
How it works
ComfyUI's advanced KSampler splits sampling so you can run it in stages - a base pass that hands unfinished, still-noisy latents to a second pass that finishes them. Two switches govern that handoff. add_noise decides whether a sampler injects fresh noise at the start (you turn it off on a second stage that's continuing from where the first left off). return_with_leftover_noise decides whether a stage stops early and passes its partially-denoised latent onward instead of cleaning it fully. Get these wrong and you either double-noise the image or finish it too soon.
This node lets you plot those choices instead of reasoning about them. XY_type picks which toggle you're sweeping - add_noise or return_with_leftover_noise - and the plot renders the on and off cases so the effect is visible. It's a small thing, but a base-to-refiner handoff going subtly wrong is exactly the kind of bug that's hard to eyeball from a single run and obvious the moment you see both states next to each other. When your two-stage SDXL result looks slightly washed out or over-noised and you can't tell why, this is a fast way to confirm the toggles are set the way you think.
The inputs and outputs
XY_type(required) -add_noiseorreturn_with_leftover_noise, the toggle to plot.
Output is X or Y (type XY), which goes to the XY Plot node's X or Y input. Because these switches belong to the advanced sampler, wire the plot into the pack's KSampler Adv. (Efficient) - the base Efficient KSampler doesn't expose them.
Installing it
ComfyUI Manager → search Efficiency Nodes for ComfyUI, install, restart. Or:
cd ComfyUI/custom_nodes
git clone https://github.com/jags111/efficiency-nodes-comfyui
Restart ComfyUI. No downloads.
Common issues
This is the one XY input where the target sampler matters. add_noise and return_with_leftover_noise are advanced-KSampler concepts, so plotting them against a basic Efficient KSampler won't do anything meaningful. Use the advanced variant.
It's also genuinely niche - if you're not building a multi-stage sample (base-to-refiner, or a manual two-KSampler split), these toggles don't come up and you don't need this node. It earns its place only in that specific workflow shape.
And the standard XY Input rule: the output does nothing until it flows through the XY Plot node into an Efficient KSampler's script input. No plot, no effect.
Pack-wide: IMPORT FAILED at startup (the recurring pip freeze non-zero exit) drops the whole pack and this node with it. Update ComfyUI, update the node, and check your Python environment.
Inputs (1)
| Name | Type | Default | Description |
|---|---|---|---|
| XY_type | COMBO | 2 options: add_noise, return_with_leftover_noise |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| X or Y | XY | — |