parameter hub 参数集中输入框

If this pack has a signature node, this is it. ParamHub (2lab) - "parameter hub 参数集中输入框", the "central parameter input box" - gathers the ten things every text-to-image workflow cares about into a single node, then fans them all back out on separate ports. One place to set your prompt, negative, dimensions, seed, steps and batch size, instead of hunting for widgets scattered across half the graph.

The reason that matters is basic graph hygiene. A shared or exported workflow with the prompt buried inside a CLIP Text Encode, the seed on a separate node, and the steps on the sampler is a workflow where people edit the wrong thing. Pull those into one hub node at the top of the graph and there's exactly one obvious place to change "what am I generating." It also fits the pack's workflow2Api story: a hub like this is precisely what a hosted API layer would read to expose the tunable parameters of a workflow.

How it works

All inputs are required, and the outputs mirror them one-for-one: prompt, negativePrompt, width, height, seed, steps, addWatermark, watermark, segment, and batchSize.

The ones you actually set every time:

• prompt and negativePrompt - multiline text, self-explanatory.
• width / height - clamped between 256 and 2048, defaults at 1024 each.
• seed - defaults to 0, and it accepts -1 if your setup treats that as "random."
• steps - default 25, capped at 1000.
• batchSize - how many images per run, default 1.

The interesting pair is addWatermark (boolean, default true) and watermark (string). In the hosted-service mindset, this is the author's watermark system - add one to every output, and the watermark field is what gets stamped. Locally, they're just outputs like the rest: wire the boolean and string wherever your workflow wants them, or ignore them. Same for segment, which looks like a grouping/billing token for the API flow and is otherwise just another string output.

The wiring pattern: hub node → width/height/batchSize/seed/steps to the KSampler and latent nodes, prompt to the CLIP Text Encode (positive), negativePrompt to the negative one. That's the whole graph's front half handled by one node.

Installing it

From AI2lab/comfyUI-tool-2lab. ComfyUI Manager → search comfyUI-tool-2lab → Install → restart, or:

cd ComfyUI/custom_nodes
git clone https://github.com/AI2lab/comfyUI-tool-2lab

Restart ComfyUI after. No models, no dependencies - pure-python, README one word long, every node labeled in 中文. One fair warning: the author never documents what addWatermark does outside his hosted flow, so on your machine it's a boolean output you're free to wire or ignore. If you build txt2img workflows meant for other people, this is the node that makes the important dials impossible to miss.