SharkOptions Start Step

This node doesn't touch your image at all. It's a settings module - one piece of RES4LYF's "options" system that you plug into a Shark or ClownsharK sampler to tell it which step of the schedule to actually start from, instead of always beginning at step zero.

Why this exists

RES4LYF's flagship node is the ClownsharKSampler, and it's built to be chained. A common pattern - you'll see it in real Wan 2.2 workflows - is running two sampler passes back to back: one for the high-noise part of the schedule with a cheaper multistep sampler like res_2m, then a second pass with a pricier substep sampler like res_2s for the low-noise finish. That only works cleanly if the second pass knows to pick up where the first one left off instead of restarting the whole schedule from scratch. That's exactly what start_at_step is for - the same job start_at_step does on the stock KSamplerAdvanced node, just wired into RES4LYF's own options ecosystem.

How the options system works

RES4LYF doesn't cram every setting onto the main sampler node - instead it has a family of SharkOptions_* nodes, each one setting a single thing (noise type, start step, and others outside this batch), with an optional options input and an options output of the same type. You chain them: plug one into the next, and the last one in the chain feeds the sampler's own options input. This is a genuine, working system - I found a real workflow question where someone was referencing a chained option by name (SharkOptions.noise_type_init) in a filename template, which confirms these are live widgets on real nodes, not decorative wiring. The catch is that none of this - the chaining pattern, what each options node does, how it interacts with the main sampler - is documented in the pack's README. You're working from the node names and your own testing.

The inputs and outputs that matter

There's really only one setting here: start_at_step (INT, default 0, range -1 to 10000) - the step index to begin sampling from. Elsewhere in this pack, -1 is used as a deliberate "leave this alone" convention (the main sampler's shift and base_shift both use -1 to mean "don't override"), so it's a safe bet the same applies here: -1 likely disables the override rather than meaning "start before step zero."

The optional options input lets you chain in whatever came before this node in your options stack; the options output is what you plug into the next SharkOptions_* node, or directly into the sampler.

How to install it

Two ways, same as the rest of the pack:

• ComfyUI Manager - search "RES4LYF", install, restart.
• Manual - activate your venv, then cd ComfyUI/custom_nodes && git clone https://github.com/ClownsharkBatwing/RES4LYF, cd RES4LYF, and pip install -r requirements.txt (or the portable Python's pip.exe if you're on the portable build). Restart ComfyUI.

The pack's fancier sampler dropdown menu (nested submenus by sampler family) needs rgthree-comfy installed alongside it - not required for this node to work, but worth having regardless since RES4LYF's author recommends it directly.

Common issues & troubleshooting

Nothing seems to happen. Check that the options output is actually wired into the sampler's options input, or into the next node in the chain. An options node sitting unconnected on the canvas does nothing - easy mistake when you're new to the pattern.

No idea what value to use. There's no documented recipe, so treat it the way you'd treat start_at_step on KSamplerAdvanced: it's the step count your previous pass ended on, which only matters if you're deliberately splitting a run across two sampler nodes.

Worried the pack is abandoned. It isn't. The original author's been quiet on public forums since mid-2025, but a second contributor has carried nearly all of the commits since January 2026, including a version bump as recently as July 2026 - active maintenance, just not from the name on the repo.