Input Parameters (Image Saver)

If the Image Saver pack has a hub node, this is it. Input Parameters bundles the six values that define a sampling run - seed, steps, CFG, sampler, scheduler, denoise - into one node with matching outputs. You wire those outputs into your KSampler and into your Image Saver. Define once, use twice, and your saved metadata is guaranteed to match what actually generated the image. That's the entire reason the pack ships loose "literal" nodes, collapsed here into a single tidy box.

For anyone who cares about accurate Civitai metadata, this is the clean way to drive a generation. No more setting CFG in the KSampler, then re-typing "7" into the saver and hoping you didn't fat-finger it.

How it works

There's no magic - it's a fan-out. Each widget you set becomes an output of the same type: the seed is an INT, the sampler is a real sampler value (the same combo type a KSampler expects), the scheduler likewise, and so on. Because the sampler and scheduler outputs are the actual ComfyUI enum values, they plug straight into a stock KSampler's inputs. Meanwhile you route the same values to Image Saver Metadata (or the full Image Saver), which records them into the A1111 params string. One source of truth, two consumers.

The inputs and outputs that matter

Six inputs, and they mirror your sampler exactly:

• seed - the generation seed.
• steps - sampling steps.
• cfg - guidance scale (default 7; note flow-matching models like Flux and Z-Image want much lower, ~1–5).
• sampler - 44 choices, from euler and dpmpp_2m through res_multistep and uni_pc.
• scheduler - the 9 core ComfyUI schedulers (simple, karras, beta, sgm_uniform, normal, and friends).
• denoise - 1.0 for text-to-image, lower for img2img.

Outputs are the same six, each typed to drop into a KSampler or a saver.

Installing it

ComfyUI Manager: search ComfyUI Image Saver, install, restart. Manually:

cd ComfyUI/custom_nodes
git clone https://github.com/alexopus/ComfyUI-Image-Saver
cd ComfyUI-Image-Saver
pip install -r requirements.txt

Restart and hard-refresh. Nothing to download.

Where people get tripped up

The scheduler list here is the core ComfyUI set. If your workflow calls for something exotic - Align Your Steps, GITS, or a scheduler that ships inside Efficiency or Inspire packs - this node won't offer it, and you'll want one of the dedicated Scheduler Selector variants instead. Same story if you're on a flow-matching model where karras actively hurts: pick the scheduler your model's docs recommend (usually beta, simple or sgm_uniform), not the SD 1.5 muscle-memory default.

The other thing: this only helps if you route the outputs to both the sampler and the saver. Feed just the KSampler and you've gained nothing over typing values in - the whole point is that the metadata reads from the same node the image did.