EasyKSampler

Here's the node that makes the whole ComfyUI-Easy-Use approach click. In bare ComfyUI, a KSampler has a wall of inputs - model, positive, negative, latent, seed, steps, CFG, sampler, scheduler, denoise. easy kSampler has almost none of that on its face, because all those settings already rode in on the pipe. You give it one cable and hit go.

That's the design: a loader packs everything into a pipe, an optional presampling node sets the seed/steps/CFG/sampler on that pipe, and easy kSampler just runs it and hands you an image. Sampling gets split from configuration, which is the pack's favourite trick - you can tweak your presampling settings and preview them without re-running the loader.

Why you'd reach for it

Because it's the natural terminal of an Easy-Use graph and it keeps that graph clean. Once your settings live on the pipe, the sampler node has nothing to clutter it, so your canvas reads as loader → presampling → sampler → done, three or four boxes instead of fifteen. It also folds the image-output step in: the node can preview, save, or hand the image to a "sender" for use elsewhere, so you don't always need a separate Save Image.

How it works

It takes the pipe - which already carries the model, conditioning, latent and all the sampling parameters set upstream - and performs the actual denoising, then decodes to an image. Because the parameters are on the pipe, this node exposes only the handful of things about output you'd set at run time. If you've used easy fullkSampler, that's the all-in-one version that holds the settings itself; easy kSampler is the split version that reads them from a presampling node, which is what you want when you're iterating.

The inputs and outputs that matter

• pipe - the one required wire, from your loader/presampling chain. This is where model, prompts, latent and settings arrive.
• image_output - what to do with the result: Preview, Save, Hide, Hide&Save, or Sender/Sender&Save (which pushes the image to a receiver node elsewhere by link_id).
• save_prefix - filename prefix when saving.

There's an optional model input if you want to override the pipe's model at the last moment (handy for applying a model patch - a PuLID or IPAdapter-modified model - right before sampling). Outputs are pipe (pass-through, so you can chain another stage) and image.

Installing it

ComfyUI Manager: search ComfyUI Easy Use, install, restart. Or cd ComfyUI/custom_nodes && git clone https://github.com/yolain/ComfyUI-Easy-Use, install requirements (install.bat / pip install -r requirements.txt), restart. No models or extra deps for the sampler itself.

Where people trip

The big one is expecting to set steps/CFG/seed here - you can't, they're not on this node. Those live on the loader or, better, on an easy preSampling node between loader and sampler. If your generation ignores a setting you thought you changed, you're editing the wrong box; find the presampling node. Second, the Sender/receiver flow throws people: it only works if a matching receiver node shares the same link_id. Otherwise, wire the image output straight into a Save/Preview and keep it simple.