Sigmas DeleteDuplicates

Simple job: Sigmas DeleteDuplicates takes a sigma schedule and strips out repeated values, so you're left with a list where every step is actually a different noise level. It's a hygiene node, not a creative one - you reach for it after some other operation in your chain leaves a schedule with dead weight in it.

Why a schedule would even have duplicates

A sampler's whole reason for stepping through a sigma list is that each step is a distinct noise level to denoise from. A step that repeats the previous value is, functionally, wasted compute - the model gets asked to do the same job twice. Duplicates tend to show up as a side effect of chaining sigma-math nodes together: concatenating two schedules that share an endpoint, clamping a bunch of values to the same floor with Sigmas Cleanup or Sigmas DeleteBelowFloor, or running a schedule through a rounding-heavy transform. None of those operations are wrong to do - they just sometimes leave duplicate entries behind, and this node is the cleanup pass for that.

How it works

It walks the input list and removes repeated entries, collapsing runs of the same value down to one. The output is the same schedule with the redundancy gone, and therefore usually a shorter list than the input - same behavior in spirit to Sigmas DeleteBelowFloor, just filtering on "is this a repeat" instead of "is this below a floor."

The inputs and outputs that matter

There's exactly one input: sigmas_1 (SIGMAS, required) - the schedule to de-duplicate. The _1 suffix is a little unusual for a single-input node (most of this pack's one-input sigma nodes just call it sigmas), but there's nothing else to wire in - no threshold, no tolerance setting for "close enough to count as a duplicate." It's an exact-match filter.

Output is a single SIGMAS list, deduplicated.

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

Your step count changes and you didn't expect it to. That's the point of the node - every duplicate removed is one fewer step in the output. If you need a fixed step count downstream, check the length with Sigmas Count after this node, not before.

No tolerance for near-duplicates. Since there's no fuzzy-match setting documented or exposed, two sigmas that are extremely close but not bit-identical (which floating-point math produces more often than you'd think) will both survive. If you're seeing duplicate-looking values still in your output, they're probably not exact matches - run the schedule through Sigmas Cleanup or round it first if you need those collapsed too.

Order is preserved. This node filters, it doesn't sort - if your input schedule wasn't monotonic going in, it won't be coming out either.