Sampler Selector

The KSampler already lets you pick a sampler from a dropdown, so Sampler Selector has the same apparent redundancy as its sibling SchedulerSelector: it's a passthrough node whose only job is exposing a choice as a standalone widget. One dropdown, all 44 samplers in ComfyUI's standard list - euler, euler_ancestral, heun, dpmpp_2m, lcm, ddim, uni_pc, and the rest - and it hands your pick to whatever you connect. Nothing more.

That's not a bug, it's the point. If you've ever dug through someone's shared workflow and lost five minutes hunting for which node actually controls the sampler (it's easy to miss inside a KSampler in a graph with forty nodes), you've felt the need for this. Pull the choice out into its own clearly-named node, label it, and now it's the first thing you see and the first thing you tweak. The author built it as a descendant of the equivalent in alexopus's ComfyUI-Image-Saver, which is the same lineage as the SchedulerSelector in this pack - the pair are meant to live together.

Inputs and outputs

One input: sampler_name, the dropdown of ComfyUI's standard samplers (44 of them, all stock - nothing custom). Two outputs:

• sampler - your choice, typed so it connects directly to a KSampler's sampler input.
• sampler_name - the same string as plain text, for logging, or for nodes that want the name as a readable STRING.

Zero compute, zero state, no math performed on anything. It's pure workflow ergonomics.

Installing it

cd ComfyUI/custom_nodes
git clone https://github.com/ARZUMATA/ComfyUI-ARZUMATA

Restart ComfyUI, or install via ComfyUI Manager by searching ComfyUI-ARZUMATA. No extra dependencies, no model files.

What it won't do for you

Set expectations before you install it: the list is ComfyUI's standard samplers only. If you're on flow-matching models and you want the RES4LYF extras - beta57, bongmath, the implicit methods, all the stuff the sampler-tuning crowd actually uses these days - this node won't give them to you. That's a different pack entirely, and the honest advice for tuning modern models is to go there instead. What this node is for is keeping your choice visible and unburied in workflows you build or share.

And the sampler you pick still has to match your model family - the node won't save you from that. On Flux, euler with a normal/beta schedule is the reliable combo (Karras is the classic trap; avoid it). For SDXL, dpmpp_2m with karras remains the safe default. This is just the widget; the sampler rules still apply.

Worth noting it's cheap insurance for anyone who shares workflows: a labeled, separate sampler selector survives someone else loading your graph far better than a setting buried in the middle of a KSampler. That's the whole reason it exists, and for that narrow job it's exactly right.