Sigmas Concat

Sigmas Concat is the straightforward stitching tool in RES4LYF's sigma toolbox, and it's worth telling it apart from its lookalike siblings: Sigmas2 Add and Sigmas2 Mult do elementwise math on two same-length schedules, producing one schedule the same length as the inputs. Sigmas Concat does something simpler - it appends sigmas_2 after sigmas_1, producing one longer schedule.

What it's actually for

This is the manual-schedule-building version of a pattern people already run in production: splitting a generation across two sampler passes. A real example from Wan 2.2 workflows - a high-noise pass with a cheaper multistep sampler like res_2m, followed by a low-noise pass with a pricier substep sampler like res_2s, split roughly evenly across the total step budget. Normally that means wiring up two separate sampler nodes. Sigmas Concat lets you build (or generate) two schedules independently - say, a coarse curve for the first phase and a fine one for the second - and hand the combined result to a single custom-sampler run instead.

How it works

Plain list concatenation: every value of sigmas_1, followed by every value of sigmas_2, in order. Nothing is reordered, smoothed, or validated for continuity between the two - that part is on you.

The inputs and outputs that matter

• sigmas_1 (SIGMAS, required) - the first schedule.
• sigmas_2 (SIGMAS, required) - the second schedule, appended after the first.

Output is a single SIGMAS list, len(sigmas_1) + len(sigmas_2) long.

How to install it

• ComfyUI Manager - search "RES4LYF", install, restart.
• Manual - activate your venv, cd ComfyUI/custom_nodes && git clone https://github.com/ClownsharkBatwing/RES4LYF, cd RES4LYF, pip install -r requirements.txt (portable builds: use the embedded pip.exe). Restart.

Common issues & troubleshooting

Order and continuity are entirely your responsibility. Nothing stops you from putting sigmas_2 first and sigmas_1 second in the wrong slots, or from concatenating two schedules that don't line up - if sigmas_1 ends near 0.05 and sigmas_2 starts back up at 0.8, you've just told the sampler to inject a fresh burst of noise mid-run. That's a real, deliberate technique in some multi-stage or re-noising setups - it's basically what happens on purpose during certain upscale or refinement passes - but it's a mistake if you didn't mean to do it. Check the actual numbers, not just that the wires are connected.

Your total step count changes. Concatenating two 15-step schedules gives you a 30-step run - obvious once you think about it, easy to forget when you're deep in a workflow. Sigmas Count on the output will confirm what you actually built before you commit a full render to it.

Not documented in the README, but the operation itself is simple enough that there's not much room for the implementation to surprise you - the risk here is entirely in what you feed it, not in what the node does with it.