CreateSampler_names ♾️Mixlab

Like LoraName, its neighbor in Mixlab's Experiment category, CreateSampler_names solves a specific, boring, genuinely useful problem: comparing several options in one queue run instead of manually swapping a dropdown and hitting generate over and over. Here the thing being compared is samplers instead of LoRAs, and the pack's README shows off the same pattern elsewhere with CkptNames, used to "compare the effects of different models" - this node is that idea applied to the sampler choice specifically.

There's exactly one input, sampler_names, a multiline STRING that ships pre-filled with every native ComfyUI sampler name, one per line - everything from euler and euler_ancestral through the DPM++ family (dpmpp_2m, dpmpp_sde, dpmpp_3m_sde), ddim, uni_pc, and the more exotic entries like res_multistep and gradient_estimation. The single output, also called sampler_names, is that same list turned into an actual list-typed output you can loop or batch through - the wildcard * typing is what lets it plug into whatever sampler-selection input expects a matching name string.

It's worth actually caring about this kind of comparison rather than treating it as busywork, because sampler choice genuinely changes output more than most people expect. Converging samplers like Euler, DPM++ 2M, DDIM, and UniPC settle into a stable image and stop changing meaningfully once you add enough steps - predictable, same seed gives the same result. Ancestral and SDE variants (Euler A, DPM++ SDE) inject fresh randomness at every step and never fully settle, so more steps changes the image rather than just refining it - more creative, less repeatable. On top of that, scheduler choice interacts with all of it: Karras schedules concentrate denoising effort in the middle of the process, which helps SD 1.5/SDXL-era models at low step counts but can actively hurt on newer flow-matching architectures that want a more even schedule. That's exactly the kind of thing that's hard to reason about from first principles and easy to just see by comparing outputs side by side - which is the entire reason this node exists.

Since this is filed under Mixlab/Experiment, treat the whole thing as a convenience utility rather than something with deep logic behind it - it's a text list turned into a list output, nothing more. If a sampler name you add doesn't work downstream, that's not this node malfunctioning; it means either you've mistyped the name or your installed ComfyUI version doesn't actually have that sampler (the list this node ships with is a snapshot of what was available at the time it was written, and newer or renamed samplers in a fresher ComfyUI build won't automatically appear here - edit the text list by hand if you need to add or remove entries).

Install follows the pack's standard route. Through ComfyUI Manager: search "comfyui-mixlab-nodes," install, restart. Manually:

cd ComfyUI/custom_nodes
git clone https://github.com/shadowcz007/comfyui-mixlab-nodes.git
cd comfyui-mixlab-nodes
install.bat

or pip3 install -r requirements.txt in a venv, or the embedded-python pip call on a portable build. No model download needed - it's a text list, not an inference node.