Nodes/ComfyUI_agilly1989_motorway/KSamplerSelect_motorway_edition
ComfyUI Node

KSamplerSelect_motorway_edition

A sampler picker for your custom-sampling pipe

By agilly1989·Created 2 years ago·Updated about a year ago· 7
KSamplerSelect_motorway_edition
  • MOTORWAY 🚌💨
  • MOTORWAY 🚌💨
sampler_name
OUTPUT_SAMPLER_keySAMPLER

KSamplerSelect_motorway_edition is a one-trick node, and the trick is fine: it wraps core ComfyUI's KSamplerSelect, which exists for the "custom sampling" style of workflow where you build a sampler stack out of separate pieces instead of using the all-in-one KSampler. You pick a sampler from the dropdown, and instead of a SAMPLER object flying out of a socket, it's stored in the Motorway under a key. That's it.

Where it fits

In a normal custom-sampling graph you chain KSamplerSelectBasicScheduler (or KarrasScheduler) → SamplerCustomAdvanced, feeding the sampler and the sigmas in as objects. In this pack the equivalent motorway-ed nodes exist for the selectors and schedulers, and the intent is that the pieces live in the pipe: this node stores the sampler under SAMPLER, a scheduler node stores its sigmas under SIGMAS, and a later SamplerCustomAdvanced-style motorway-ed node picks both up by key. The single widget here, sampler_name, is the standard dropdown of 34 samplers - euler, euler_ancestral, heun, dpm_2, dpmpp_2s_ancestral, dpmpp_2m and friends.

Worth knowing: the MOTORWAY 🚌💨 input is optional on this one, because there are no non-widget inputs to pull from the pipe. Leave it unconnected and it still works - it just writes the SAMPLER key and moves on. So it doubles as a standalone sampler picker that happens to be pipe-friendly.

Which sampler should you pick? The dropdown itself doesn't care, and the KB's guidance transfers unchanged: converging samplers (euler, dpmpp_2m, ddim) settle to a stable image, ancestral ones (euler_ancestral, dpmpp_2s_ancestral) keep injecting noise and stay more creative but less predictable. DPM++ 2M Karras remains the safe default for SD 1.5/SDXL; newer flow-matching models usually want their own recommendations.

Set it and forget it

The only things you set:

  • sampler_name - your sampler.
  • OUTPUT_SAMPLER_key (default "SAMPLER") - the motorway key the chosen sampler lands under.

Everything downstream that wants a sampler looks it up by that key. Keep the default unless you already have something stored under SAMPLER that you'd overwrite.

Install and reality check

cd ComfyUI/custom_nodes
git clone https://github.com/agilly1989/ComfyUI_agilly1989_motorway

Restart ComfyUI, or install via ComfyUI Manager (search "motorway"). No pip deps, no models to download.

One honest note: this is a trivial wrapper, and the pack around it is an active-development beta. The current repo HEAD has the generated _motorway_edition nodes commented out of registration - a hotfix mid-refactor - so if the node simply isn't in your menu after install, that's expected of this pack right now, not a sign you installed wrong. And if a run dies with 'MotorwayClass' object has no attribute 'hash_', you typed a key that was never written to the pipe. For a node this small, if you're not already committed to a motorway workflow, the stock KSamplerSelect costs you nothing extra - this one only pays off inside the pipe.

Categoryagilly1989 Nodes/Motorway-ed/sampling/custom_sampling/samplers

Inputs (3)

NameTypeDefaultDescription
sampler_nameCOMBO34 options: euler, euler_cfg_pp, euler_ancestral, euler_ancestral_cfg_pp, heun, heunpp2, +28
OUTPUT_SAMPLER_keySTRINGSAMPLER
MOTORWAY 🚌💨optMOTORWAY 🚌💨

Outputs (1)

NameTypeDescription
MOTORWAY 🚌💨MOTORWAY 🚌💨