🐧 Sigmas Graph

A sigma schedule is a list of numbers, and lists of numbers hide problems. A step that's way out of line, a tail that drops too hard, a schedule that's really only doing two thirds of what you think - all of it is invisible until the render comes out wrong and you start counting steps by hand. This node makes the schedule visible: it plots the sigma curve and shows it right in the node, so you can see the shape of your denoising before you spend GPU time on it.

It's part of the 🐧 SuperNodes pack by SuperCC, and it's the pack's scheduling family's debug display.

How it works

One input, sigmas - any SIGMAS output, typically from a scheduler or from a chain of the pack's Sigma Insert / Replace / Remove / Rescale nodes. The node plots the curve with matplotlib (that's the pack's one extra dependency, and this is the only node that uses it): dark background, blue line, markers on every step, step index on the x-axis and sigma value on the y-axis. It's an output node, so it just displays - wire it anywhere convenient and read the picture.

The output is also a real IMAGE tensor of the plot, so if you want the graph saved into your output folder alongside your render - useful for documentation or for comparing two schedule versions side by side - you can wire it to a Save Image node.

When you'd reach for it

• Debugging a custom schedule: you set it up by feel, and you want to see where the steps concentrate before committing to a long render.
• Verifying edits: insert, remove, replace, rescale - then graph and confirm the shape is what you intended.
• Learning. If you've never actually looked at what Karras vs. linear vs. your custom scheduler produce as curves, this node is the fastest way to build the intuition. The difference between "middle-heavy" and "front-loaded" stops being a phrase and becomes a picture.

Install

ComfyUI Manager, search "ComfyUI-SuperNodes", or:

cd ComfyUI/custom_nodes
git clone https://github.com/sonnybox/ComfyUI-SuperNodes

Restart ComfyUI. No model downloads. This is the node that needs matplotlib - ComfyUI Manager installs the pack's requirements automatically, but if you cloned manually and skip pip install -r requirements.txt, the graph node is the one that breaks. Also: the pack uses ComfyUI's newer extension API, so keep ComfyUI updated or nothing registers.

Gotchas

• If the node errors on import of matplotlib, that's the missing dependency - install it (pip install matplotlib in your ComfyUI environment) and restart.
• It's a visualization of the schedule tensor as-is, including the terminal 0.0. A schedule that "looks like a cliff at the end" is real information, not a rendering artifact.
• It adds a little startup cost per graph render (matplotlib import is slow the first time). Negligible on a per-render basis.

It's a debug tool, not a workflow component, and that's fine. Every other sigma node in this pack gets twice as useful the moment you can actually see what they're doing to your curve.