AK KSampler Settings

If you've ever run a workflow with four KSamplers and wanted to change the seed or the CFG in all of them at once, AK KSampler Settings is the answer: it's a single node that holds every sampling parameter in one place, packs them into a string, and lets its partner node (AK KSampler Settings Out) unpack them wherever you need them. Change the settings once, every sampler downstream follows.

This is one of the "interesting parts" the author actually calls out in the README, alongside Project Settings and AK Base - it's the backbone of his whole large-workflow setup, where six to eight samplers run simultaneously and you don't want to hunt through the graph for a single CFG slider.

What it holds. The settings object contains exactly what a KSampler needs: seed, steps, cfg, sampler_name, scheduler, denoise - plus one extra field, xz_steps, which is specific to the author's sister pack ComfyUI-AK-XZ-Axis (an XY-plot tool that works with any KSampler). xz_steps controls how many steps each plot cell runs, and it's the field AK Base automatically drops to 1 when you select a single image in its gallery. If you don't use XZ-Axis, you can leave it alone.

The inputs are the full ComfyUI lists - 44 samplers and 9 schedulers, so euler, dpmpp_2m, karras, exponential and everything else from your core ComfyUI is there. Note the seed input is deliberately named seed_value, not seed. That's on purpose: ComfyUI can auto-randomize an input literally named seed depending on your global seed behavior, and the author didn't want the settings node to look "changed" every run. So the node keeps control of its own seed.

The one output. sampler_settings, a STRING. It's a compact JSON blob (with a checksum hash and the node's ID baked in so downstream can detect changes). Don't read it - feed it to AK KSampler Settings Out.

How to install. ComfyUI Manager, search "AK Pack", or:

cd ComfyUI/custom_nodes
git clone https://github.com/akawana/ComfyUI-Utils-extra

Restart ComfyUI. No dependencies, no models.

Where people get burned. The classic mistake is expecting this node to be the sampler. It isn't - it only produces a settings string; you always need AK KSampler Settings Out on the receiving end to turn that string into the typed values a real KSampler accepts. Also, because settings flow through a string, there's no type checking until the Out node parses it - which is fine when both nodes come from this pack, but don't try to hand-build the JSON yourself. And if you're not using AK Base or XZ-Axis, the xz_steps field is dead weight you can ignore; the rest of the node is still worth it just for centralizing your seeds.