Scheduler Selector (for KSampler (Efficient))

What this node actually is

If you landed here because you found a node called "Scheduler Selector" inside a Telegram pack and felt a small moment of confusion - your instincts are working. This node has nothing to do with Telegram. It's a tiny utility that Sergey004 bundles in his ComfyUI-Telegram-Sender repo (the category is literally Telegram/utils), and it does one simple job: it's a dropdown that hands you back a scheduler name as a string.

A scheduler, if you're new to this, is the noise schedule a sampler walks through - karras, normal, sgm_uniform, and friends. Different schedulers shape how denoising progresses, and a schedule that flatters one model can dull another. Normally you pick it from the KSampler's own dropdown. This node gives you that same choice as a standalone widget, plus a few options a stock KSampler doesn't offer.

Why you'd reach for it

Two genuinely useful things:

1. One control for many samplers. Wire the output into several KSamplers and you can switch the whole workflow's schedule from a single node, instead of clicking through every sampler. Handy in big multi-pass graphs where you want everything on the same curve.
2. Extras the stock list lacks. The node appends AYS SD1, AYS SDXL, AYS SVD, and GITS to ComfyUI's built-in scheduler list. AYS ("Align Your Steps") presets are tuned step-count/schedule combos, and the community verdict on them is genuinely positive - low-step benefits with no real downside for most models.

The display name says "for KSampler (Efficient)" because the author built it to feed the popular efficiency-nodes all-in-one sampler. It works with anything that takes a scheduler input, though.

How it works

The whole implementation is about ten lines. At load time it reads comfy.samplers.KSampler.SCHEDULERS - whatever scheduler names your ComfyUI build ships - tacks on the four extras, and exposes that list as the choice. When the graph runs, it returns the string you picked, unchanged. No math, no conversion, no hidden behavior. It is a pass-through with nicer packaging.

Inputs and outputs

There's exactly one of each, so this is the rare node where you can't get lost:

• scheduler (required) - a dropdown with 13 options: simple, sgm_uniform, karras, exponential, ddim_uniform, beta, normal, linear_quadratic, kl_optimal, AYS SD1, AYS SDXL, AYS SVD, and GITS.
• scheduler (output) - the same value as a string, ready to wire into a sampler's scheduler input.

Installation

It ships inside the pack, so installing the pack is installing this node. Via ComfyUI Manager: "Install Custom Nodes" → search ComfyUI-Telegram-Sender → install → restart. Or manually:

cd ComfyUI/custom_nodes
git clone https://github.com/Sergey004/ComfyUI-Telegram-Sender
pip install requests pillow piexif

Then restart ComfyUI. No model files to download; the dependencies are lightweight.

Common issues

The one trap: you can't just drag this output onto a stock KSampler. A KSampler's scheduler is a widget, not an input, so you have to right-click the KSampler and choose Convert scheduler to input first. Miss that step and there's simply no socket to plug into.

Second, an honest caveat: the pack's README warns it "can be broken on latest versions," and this node inherits that risk. Installing a whole Telegram pack just to get a scheduler dropdown is a lot of luggage - the pack also fires a background thread at startup that tries to fetch Civitai previews for your models. If you already have the pack for Telegram, this is a nice free utility; if you don't, there are lighter ways to get AYS/GITS into a dropdown.