Sampler Params

This is the node that makes the rest of this pack's sampler tooling work: Sampler Params takes the six settings you'd normally scatter across separate sockets - sampler, scheduler, steps, denoise, seed, and cfg - and bundles them into a single sampler_params value that travels over one wire.

The whole point is reuse and portability. In vanilla ComfyUI, every sampler node is a fresh six-socket negotiation, and if you run multiple sampling passes in one graph (a detailer pass, a hires pass, a second pass with different denoise), you're re-entering or rewiring the same numbers over and over. Bundle those settings once, and you can pass the same configuration to the pack's SamplerCustom (Sampler Params) node, store it in image metadata via Set Sampler Params Extra, or override individual fields downstream with Sampler Params Context.

The inputs

All six are plain widget fields - you type them, nothing to wire:

• sampler (default euler) - the sampler name. The dropdown is populated directly from ComfyUI's own sampler list, so anything you'd see on a stock KSampler is available: euler, euler_ancestral, dpmpp_2m, dpmpp_sde, uni_pc, and friends.
• scheduler (default simple) - same story, drawn from ComfyUI's scheduler names, so karras, exponential, sgm_uniform and the rest are there.
• steps (default 20, 1–10000)
• denoise (default 1.0, 0.01–1.0)
• seed (default 0) - a plain integer. One thing worth knowing: this node doesn't attach a "randomize after run" behavior the way a stock KSampler widget does, so if you want a fresh seed every run you'll want to drive it from elsewhere. The pack ships a Seed Generator node for exactly that.
• cfg (default 7.0, 0–100)

The single output, sampler_params, plugs into SamplerCustom (Sampler Params) - or into Split Sampler Params if you bundle first and need the values loosened up again.

One thing to calibrate

Because the defaults here (euler + simple + cfg 7) are generic, don't treat them as a recommendation. The right combo depends heavily on your model family: SD 1.5/SDXL shrugs happily at DPM++ 2M + Karras; Flux wants euler with a non-Karras scheduler and guidance handled separately; flow-matching models like Anima expect their own schedulers. The bundle is a container - it doesn't know or care what's inside, and it won't correct a sampler that's wrong for your checkpoint. Set the values for your model, then let the bundle do the bookkeeping.

Install

Same pack, same drill:

cd ComfyUI/custom_nodes
git clone https://github.com/kinorax/comfyui-info-prompt-toolkit.git
cd comfyui-info-prompt-toolkit
pip install -r requirements.txt

Or ComfyUI Manager → search "ComfyUI-Info-Prompt-Toolkit" → install → restart. No models, no extra dependencies beyond the pack's base requirements.txt.

Where people get burned

Mostly by expecting the bundle to be magic. It isn't - garbage in, garbage out, and the defaults will happily produce mediocre images if your model wants something else. Also worth remembering: the bundle is an IPT-SamplerParams typed value, not a plain string. You can't type it into a text field or pass it to a node that expects a loose value - that's what the Split node is for. If you're building toward the pack's metadata round-trip (save an image, reload its settings later), make sure this bundle feeds the sampler rather than a hand-wired duplicate, or your saved infotext will carry settings you're not actually using.