Sigmas DotProduct

A dot product is normally a reduction - you take two vectors of the same length and get back a single number, the sum of their elementwise products. Sigmas DotProduct takes two sigma schedules and, per its type signature, hands you back another SIGMAS list rather than one scalar. That means it isn't computing a textbook dot product and stopping there; something in the implementation is either doing a running or elementwise version of the operation, or broadcasting a single scalar result back out across a list. I'll say upfront: there's no README section, source comment, or community thread that spells out which, so treat the exact mechanism as an educated read of the node's type signature rather than a documented spec.

What I can tell you with confidence

Two SIGMAS schedules go in, one SIGMAS schedule comes out, and there's a normalize toggle. That last part is the more informative clue - normalizing before a dot product is the standard move for computing cosine similarity (how aligned two vectors are, independent of their magnitude) rather than a raw magnitude-sensitive dot product. So this node most plausibly measures or blends how similar two schedules' shapes are, not just their raw values, when normalize is on.

The inputs and outputs that matter

• sigmas_a (SIGMAS, required) and sigmas_b (SIGMAS, required) - the two schedules to combine.
• normalize (BOOLEAN, default false) - per the note above, likely switches between a raw-magnitude combination and a shape/direction-focused one.

Output is a single SIGMAS list.

Should you use this?

Honestly, only if you enjoy poking at things without a manual. I checked the community record for any discussion of RES4LYF's vector-math sigma nodes - this one, Sigmas CrossProduct, the arc-trig reshaping tools - and found nothing: no workflow shares, no forum threads, nobody reporting a recipe that uses it. That's consistent with the rest of this corner of the pack: RES4LYF ships a large pile of raw mathematical primitives for building custom schedules by hand, and not every primitive in that pile has a known best use. If you want a predictable way to blend two schedules, Sigmas Lerp (proportional blend) or Sigmas2 Add/Sigmas2 Mult (elementwise combine) have much clearer, documented-by-implication behavior. This one is for experimenting.

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

Mismatched schedule lengths are undefined territory, same caveat as the pack's other two-schedule math nodes - build sigmas_a and sigmas_b from schedulers set to the same step count to sidestep the question entirely.

Don't assume the output is monotonic or non-negative. Whatever this node computes, there's no reason to expect the result respects "should decrease smoothly toward zero" the way a sampler wants. Run it through Sigmas Abs and Sigmas Cleanup, and check the actual values with a preview node, before it goes anywhere near a render you care about.