Sampler Selector

The Sampler Selector looks like a waste of a node - one dropdown, one output, nothing else. It's not. It's the difference between choosing a sampler in five places and choosing it in one.

Stock ComfyUI buries the sampler as a widget inside every KSampler. Change your mind and you're clicking through a 60-item dropdown in each node, and nothing downstream can ever read which sampler you used. This node pulls that choice out onto a wire. Pick "dpmpp_2m" once, and the value flows into your KSampler and into HelperNodes_SaveImage, which records it in the image metadata. One decision, whole graph.

How it works

It's a pure passthrough. The dropdown is populated from comfy.samplers.KSampler.SAMPLERS - the exact same list the stock KSampler shows. The node adds nothing, filters nothing, and doesn't even validate your taste. You select sampler_name, it hands back the same value as the sampler output. That's the whole mechanism, and honestly it's the right one: a selector that duplicated the list would rot the moment ComfyUI shipped a new sampler.

One thing worth knowing: the list tracks your ComfyUI version. The build this was measured against exposes 34 samplers, including the newer cfg_pp family and gradient_estimation. If your dropdown looks different, that's your ComfyUI, not a broken install.

Inputs and outputs

• sampler_name - the one required input, an enum of everything KSampler knows.
• sampler output - same list, typed as the sampler enum. Wire it into KSampler's sampler and Save Image's sampler_name.

That's it. There's no optional anything.

The honest take

This node doesn't pick a sampler for you, and it won't fix your defaults. The old "DPM++ 2M + Karras" safe bet is not safe anymore: it belongs to DDPM-style models like SD 1.5 and SDXL, while flow-matching and distilled models want Euler-family samplers on conservative schedules, often at far fewer steps. And no, the list won't include RES4LYF solvers - if you want those, install that pack and choose them here like anything else. What the selector buys you is the habit of making the choice once, on a wire, where it can be logged and shared.

Install

ComfyUI Manager - search ComfyUI-Helper-Nodes, install, restart. Or manually:

cd ComfyUI/custom_nodes
git clone https://github.com/teward/ComfyUI-Helper-Nodes

Then restart ComfyUI. The pack's requirements.txt lists numpy, pillow, piexif and whratio; numpy and pillow already ship with ComfyUI, and Manager handles the rest. No model files to download - this is pure utility.

Troubleshooting

If the node is missing after install, restart is the usual fix. If the sampler names don't match a tutorial you're following, check the ComfyUI version first - and if a wire isn't connected, the choice does nothing. A selector with no wire is just an ornament.