EZ Sampler

EZ Sampler is the settings node that replaces the entire sampler side of a stock workflow - the KSamplerSelect, the scheduler, the sigma plumbing - with four fields. It packages your sampling choices into a settings bundle, and EZ Generate does all the node-wiring when it runs. You configure four things and never look at a BasicScheduler again.

• sampler_name (combo, default euler) - the sampling algorithm. It's populated from ComfyUI's actual sampler list, so everything you'd find on a stock KSampler is here: euler, euler_ancestral, dpmpp_2m, dpmpp_2m_sde, dpmpp_sde, uni_pc, and friends. If you don't have a strong opinion, euler (the default) or dpmpp_2m are the sensible starting points; _ancestral variants add variety per step at slight quality cost.
• scheduler (combo, default normal) - how noise is removed across steps. normal covers the standard Karras-ish behavior most people mean by "default." The pack also exposes three specials: align_your_steps (for SDXL), gits, and sdturbo - these map to dedicated scheduler nodes (AlignYourStepsScheduler, GITSScheduler, SDTurboScheduler) under the hood, so the dropdown genuinely reflects real behavior, not branding.
• steps (INT, default 30, range 1–10000) - denoising steps. 30 is a fine all-purpose default for SDXL-lineage; distilled or turbo-style setups want far fewer.
• seed (INT, 0 to 2^64-1) - the noise seed, with a control-after-generate randomize toggle so you can roll new seeds between runs with one click, the same way you would on a stock KSampler.

The output is SAMPLER (EZ_SAMPLER_SETTINGS), wired into EZ Generate's sampler input.

What's actually going on

EZ Sampler stores your four values. At generation time EZ Generate builds the sampler graph: a KSamplerSelect, the scheduler node matching your choice, a CFGGuider (CFG comes from your EZ Prompt weight, remember), and a SamplerCustomAdvanced - all strung together internally. So the node is a config object, not a computation; nothing happens until EZ Generate executes, and then it happens exactly as you configured it.

That's the trade to understand: you give up seeing the individual sampler nodes on the canvas, and in exchange you get a workflow that's genuinely readable. For someone iterating on styles, the four fields are where the dials should be - sampler and scheduler are two dropdowns, not a thicket of wiring.

One honest note on defaults

The default scheduler being normal and 30 steps is a safe, slightly conservative baseline. If you're generating with the pack a lot, the biggest quality-per-click wins come from experimenting with scheduler rather than sampler - Karras-style schedules usually edge out normal for detail. The special schedulers exist for a reason; sdturbo, for instance, is for the few-step turbo-lineage models and will behave badly at 30 steps, so read the model's docs before selecting it.

Installing it

It ships with the Prompt Helpers pack. In ComfyUI Manager search "Prompt Helpers" (Pauan/comfyui-prompt-helpers), install, restart. Manual:

cd ComfyUI/custom_nodes
git clone https://github.com/Pauan/comfyui-prompt-helpers

Restart ComfyUI. No model downloads; the pack needs PyYAML and desktop-notifier and ComfyUI 0.8.0+. The special schedulers ship with ComfyUI core, so there's nothing extra to fetch.

If a render comes out noisy or flat, the usual culprit is a scheduler/steps mismatch (turbo scheduler at high step counts, or a normal scheduler with too few steps) - not the node. Reset to the defaults and change one variable at a time.