basicIn_Scheduler

Same idea as this pack's basicIn_Sampler, applied to the other half of that pairing: instead of a scheduler dropdown buried inside every KSampler node in your graph, you get one standalone node with the choice, and an output you fan out to as many samplers as need it. If you're running a multi-stage pipeline and want every sampling pass to use the same noise schedule without manually syncing several dropdowns, this is the node for that.

What it is

One dropdown, scheduler, with all 9 of ComfyUI's built-in schedulers: simple, sgm_uniform, karras, exponential, ddim_uniform, beta, normal, linear_quadratic, kl_optimal. One output, scheduler, matching. That's the entire node - a router, same as its sampler counterpart.

The scheduler choice actually matters, and it's model-dependent

This is the one place where getting it wrong genuinely costs you image quality, not just tidiness. Karras concentrates noise removal in the middle of the sampling trajectory, which is a real improvement for older DDPM-style models like SD 1.5 and SDXL - it's why "DPM++ 2M Karras" became the default combo people reach for. But on flow-matching architectures (Flux, Z-Image, Klein, Anima - most of what's current), that same mid-trajectory concentration is a distortion, not a correction, because those models were trained on a near-straight noise-to-image path that doesn't benefit from reshaping. A 62-sampler by 16-scheduler community sweep on Z-Image Turbo landed on a blunt one-line summary: avoid Karras and Exponential specifically, and reach for beta, simple, or linear_quadratic instead. If you're getting muddier results after switching model families and haven't touched your scheduler, this is the first thing to check.

Pair it with basicIn_Sampler, not on its own

Sampler and scheduler are always a pair in practice - neither choice means much without the other, which is why this pack ships them as two separate but obviously matched nodes (basicIn_Sampler handles the other half). If you're setting up a multi-stage pipeline, it's worth wiring both standalone nodes side by side near the top of your graph rather than one now and the other later, so anyone reading the workflow later sees the pairing at a glance instead of hunting for a scheduler choice buried three nodes downstream.

Installing it

Search "ComfyUI-Apt_Preset" in ComfyUI Manager, or:

cd ComfyUI/custom_nodes
git clone https://github.com/cardenluo/ComfyUI-Apt_Preset.git

Run install.bat on Windows (pip install -r requirements.txt on Linux/Mac), restart. No models involved - pure dropdown routing.

Common issues

Same caveat as its sampler sibling: this node routes a name from ComfyUI's own registry, so a missing option almost always traces back to a ComfyUI core version mismatch rather than a bug here. And worth repeating because it's the single most common mistake people carry over from older SDXL-era habits: don't assume Karras is a universally safe default just because it used to be. Check what architecture the model you're running is actually built on - DDPM-style versus flow-matching - before picking a scheduler out of muscle memory.