Essential Params

Essential Params collects the five sampler settings you fiddle with on every single generation - CFG, steps, sampler, scheduler, denoise - into one node and hands them out as a single "EParams" packet. You set them once, in one tidy place, and everything downstream that needs sampling parameters pulls from that one source of truth.

The payoff is workflow hygiene. Normally a KSampler buries its settings inside itself, and if you want the same sampling settings across two samplers (a hires pass and a base pass, a batch of variations), you're either setting them twice and hoping you remember, or dragging wires from primitives. With a params packet, changing "20 steps" in one node updates every consumer at once. It's the same pattern the ecosystem has converged on - bundle parameters, expand them where needed - and it's why this pack's sampler-adjacent nodes expect an EParams input.

How it works

The node stores its five values as a tuple and hands that tuple out through a single custom-type socket (EParams). On its own it doesn't do anything else - no sampling happens here. The magic is on the other side: the pack's Essential Params Expand node unpacks the packet back into individual outputs (cfg, steps, denoise, sampler, scheduler, an ays flag, and even a ready-made SAMPLER object) that wire straight into a KSampler or a custom sampler node. The two are a matched pair - you almost never use one without the other. It's openly inspired by the sv-nodes family, per the source's own comment.

Inputs and outputs

The five knobs:

• cfg (FLOAT, 0–20) - guidance scale. Note the default is 1, which is the right starting point for guidance-distilled models (Flux-style and everything Turbo/Lightning-ish), and a wrong starting point for SD 1.5/SDXL where 5–9 is typical. Know which family your checkpoint is in before you trust the default.
• steps (INT, 1–100) - denoising steps. Default 20.
• sampler - every sampler ComfyUI knows, from euler and euler_ancestral through dpmpp_2m, ddim, and the whole modern roster.
• scheduler - the standard schedulers plus ays, the Align Your Steps option.
• denoise (FLOAT, 0–1) - how much of the image to regenerate. 1 = full generation, lower = keep more of the input.

Output:

• _ (EParams) - the packet. A bit of an ugly name, but it's just the bundle.

Installing it

Ships in ComfyUI Fictiverse Nodes. ComfyUI Manager → search "ComfyUI Fictiverse Nodes" → install → restart, or:

cd ComfyUI/custom_nodes
git clone https://github.com/Fictiverse/ComfyUI_Fictiverse

No extra dependencies - it imports comfy.samplers, which is core. No models. Apache 2.0.

Common issues

Pack gotcha: ModuleNotFoundError: No module named 'custom_nodes.ComfyUI_Fictiverse' means the folder isn't named exactly ComfyUI_Fictiverse; rename and restart.

For this node, the big gotcha is the CFG default of 1. If you're on SDXL or Pony/Illustrious and you just click run, CFG 1 gives you washed-out, under-guided images - the classic "why is this so bland" moment. And there's a second layer: the packet is only meaningful to nodes that understand EParams. You can't plug it into a stock KSampler - that needs individual inputs, which is exactly what Essential Params Expand is for. If a workflow you downloaded has an EParams wire going nowhere, the sampler will just use its own defaults.