Schedule Comparison

The Schedule Comparison node does one thing and does it carefully: it takes two verified Sigmax schedules and compares them, by terminal-inclusive sigma index, into a comparison_report string. No images, no sampling, no judgment about which schedule is "better" - just a precise, index-aligned diff of the noise-level curves.

If that sounds dry, here's why you'd actually want it. When you're tuning a flow-matching workflow, the interesting question is often not "what does my new schedule look like" but "how far did I actually move from the old one" - at step 4, is the new curve denoising harder or gentler? Schedule curves are just lists of numbers, and eyeballing two lists doesn't work. This node aligns them by terminal-inclusive index (so the terminal/zero sigma lands at the same position on both sides) and reports the difference explicitly.

Inputs and output

• sigmas_a / schedule_info_a - the first schedule and its info JSON, straight from any Sigmax scheduler.
• sigmas_b / schedule_info_b - the second.

The output is a single comparison_report string. Read it; it's the whole deliverable.

Both inputs say "verified" for a reason. This pack's schedule nodes each emit a schedule_info JSON alongside the sigmas, and the inspection family uses that metadata as the verification key - the comparison is only meaningful when both sides are schedules the pack can place on its versioned, fingerprinted footing. That's also your honesty check: if a schedule didn't come from Sigmax, this node isn't going to make claims about it.

Installing

Same pack, same one-liner:

cd ComfyUI/custom_nodes
git clone https://github.com/rookiestar28/ComfyUI-Sigmax comfyui-sigmax

Restart ComfyUI and search Sigmax; ComfyUI Manager lists "ComfyUI-Sigmax". Python 3.10+, ComfyUI 0.29.0+, no extra dependencies.

Where it fits, and where it doesn't

It fits the experimental loop: build a schedule with the AdvancedFlowMatchScheduler, tweak the shift, compare against the previous one, and see exactly what changed before you spend GPU time sampling both. It doesn't fit "which of these renders better" - that's a sampler-side question and this node will not answer it. And a practical note: the node expects Sigmax schedules. Comparing arbitrary external sigmas is out of its contract, so don't feed it a random BasicScheduler output and expect a meaningful report.