Sampler Config

Sampler Config is the pack's biggest swing: it replaces the front panel of a KSampler - seed, steps, CFG, sampler, scheduler, denoise, and even the empty latent - with a single node that hands all of it out on separate outputs. "One node for most KSampler inputs," as the README puts it.

What it actually does

It's a config hub plus an Empty Latent in one package. You set everything on this one node, and it emits ten outputs: Empty_Latent, Seed, Steps, CFG, Sampler, Scheduler, Denoise, Width, Height, Batch_Size. Wire those straight into a KSampler's matching inputs and the sampler's own widgets become irrelevant. Change one setting in one place and it propagates.

The layout makes sense if you're building a "control panel" workflow - all your tunables in one box instead of scattered across the graph. It also centralizes the aspect-ratio logic: pick a preset like "16:9 - 1344 x 768" and the node resolves Width and Height for you.

The settings that matter

• Aspect_Ratio - the headliner. Eighteen presets from "1:1 - 512 x 512" up to "21:9 - 1536 x 640", plus Custom. Pick a preset and it overrides Width/Height entirely. The tooltip on Width/Height says it plainly: "This only matters if you picked a custom Aspect Ratio."
• Width / Height - used only when Aspect_Ratio is Custom. Both default to 1024.
• Sampler / Scheduler - the full stock ComfyUI lists (44 samplers, 9 schedulers) straight from comfy.samplers, so everything your ComfyUI build knows about is available here.
• CFG, Steps, Denoise, seed, Batch_Size - standard KSampler values, with sane defaults (30 steps, CFG 8, denoise 1.0).

The generated Empty_Latent is exactly what ComfyUI's own EmptyLatentImage produces - a zero-filled tensor at batch × 4 × height/8 × width/8, sized from your final Width/Height. That's the one genuinely hidden bit of behavior: this node is your latent source.

Where people get burned

The latent trap: because it hands you an Empty_Latent, beginners assume the node is optional. It isn't - if you want its size logic to do anything, the Empty_Latent output has to feed your sampler. Wire it.

Second, the aspect presets are SDXL-era sizes. If you're generating on a model that wants 1024+ across the board (or a very different native resolution), the small presets will produce small images; stick to Custom and set your own numbers.

And remember the sampler/scheduler advice hasn't changed: the "right" pair depends on your model family. DPM++ 2M Karras is an SDXL-era default, not a universal law - flow-matching models (Flux and friends) want Euler-family samplers on conservative schedules, and distilled models run CFG at 1. Sampler Config just exposes the choice; it doesn't make it for you.

Installing it

ComfyUI Manager → search "HavocsCall" → install → restart. Manual:

cd ComfyUI/custom_nodes
git clone https://github.com/HavocsCall/comfyui_HavocsCall_Custom_Nodes

Restart, under HavocsCall/Basic. No requirements.txt, no model downloads - instant.