Legacy2_ClownsharKSamplerOptions

Same story as its sibling ClownsharKSamplerAutomation_Advanced: this shows up as Legacy2_ClownsharKSamplerOptions in older workflows, kept around because RES4LYF's README promises old node names still work after the rewrite, just prefixed. It bundles the pack's noise-generation settings - the stuff the README documents under "NOISE SETTINGS" for the current all-in-one sampler - into a standalone node that outputs one OPTIONS object instead of living as widgets on the main sampler. If you're building new, use the current sampler's own fields; this is for keeping old graphs alive.

How it works

No sampling happens here either - it's validation and packaging. You set a pile of noise-related values, it hands them off as an OPTIONS object that the legacy ClownsharKSampler reads.

The inputs and outputs that matter

There's a lot on this node, but most of it you'll never touch. The ones that actually change your output day to day:

• noise_type_init and noise_type_sde - which noise pattern seeds the initial latent versus which one gets used during SDE sampling steps. Both are 17-option enums (gaussian, fractal, uniform, and a family of pyramid-* variants among others). Default is gaussian / brownian.
• noise_mode_sde - how the eta value below scales into actual noise amount. Six choices (hard, hard_var, hard_sq, soft, softer, exp), listed in the README in order of how strong the effect is. exp is the odd one out - it's the only mode that tolerates values well above 1.0 without triggering internal scaling to avoid NaNs.
• eta (default 0.25) - the main dial. Controls how much noise gets injected after each SDE step.
• noise_seed (default -1) - works like the core KSampler's seed widget; -1 increments off the last one used.

Everything else is deeper-end tuning: s_noise and d_noise are extra scaling multipliers on top of eta; alpha_init/k_init and alpha_sde/k_sde shape the pyramid-family noise types specifically and aren't something you tune by feel; c1, c2, c3 are raw custom coefficients; and t_fn_formula / sigma_fn_formula are literal formula strings for defining your own time/sigma functions - this is for people writing their own solver math, not a knob to poke experimentally. There's also an optional options input if you want to chain in and extend an existing OPTIONS object rather than build one from zero.

Output: options (OPTIONS), wired into the legacy sampler.

How to install it

Via ComfyUI Manager (search "RES4LYF"), or:

cd ComfyUI/custom_nodes
git clone https://github.com/ClownsharkBatwing/RES4LYF/
cd RES4LYF
pip install -r requirements.txt

then restart.

Common issues & troubleshooting

Don't feel obligated to touch most of these fields. The defaults are sane; the two that actually move the needle for most people are eta and the noise_type_* pair. Everything past that is for people chasing a specific effect who already know what they're chasing.

A malformed t_fn_formula or sigma_fn_formula is the most likely source of a raw Python traceback rather than a normal ComfyUI red-link error, since these are evaluated as expressions. Leave them blank unless you're deliberately writing custom solver math.

If you didn't inherit this from an old workflow, you probably don't need it at all. The current all-in-one ClownsharKSampler exposes the same noise settings directly on the node - no separate options object to wire up.