People Evaluation Node (Creepybits)

People Evaluation Node is the pack's most personality-heavy "Creative Council" member. It's a dropdown of four canned personas - attractiveness_nice.txt, attractiveness_rude.txt, attractiveness_x.txt, attractiveness_xx.txt - and its whole job is to load one of those instruction files and hand you the text. Like its sibling Evaluater Node, the "evaluating" happens downstream, in whatever LLM you feed the text to.

The point is less practical than the rest of the suite and more, let's say, experiential. You generate an image of a person, pipe it into Gemini with one of these personas as the system prompt, and get back a verdict with the chosen attitude: constructive and kind, bluntly critical, or increasingly unhinged depending on which file you picked. It's a novelty node with a serious edge - the _x and _xx variants are not polite, and they're the kind of content that makes Gemini's safety filters decide today is not the day.

How it works

Structurally identical to Evaluater Node: at startup it reads evaluate_people.json in the pack's assets folder, which lists the four prompt files. The prompt_file dropdown is built from that list. Running the node loads the selected .txt from assets/prompts and outputs it as text. There's an optional_input wildcard socket, but the shipped code accepts it and ignores it - don't expect it to change anything.

The inputs that matter

Just prompt_file. Pick a persona, run, and take the text output to the system_prompt input of the pack's Gemini node, with your image on Gemini's image input. That's the whole recipe. There's no image input on this node itself - it's a text loader, nothing more.

Two things to know before you wire it up

First, the caveats that apply to every persona node: no Gemini downstream, no evaluation. And the _x/_xx personas will reliably run into Gemini's safety settings at the default thresholds, returning error strings instead of verdicts. If you specifically want the spicy variants to work, the pack's Gemini node's safety_threshold input has a "Block None" setting for a reason.

Second, a content note worth stating plainly: this node exists to have an AI score people's attractiveness. It's a toy, and the "evaluator" is a language model confidently making subjective judgments. Take the output accordingly - it's for fun, not for feedback you should act on. The pack ships it, so it's in your menu either way; just know what you're reaching for.

Installing it

Part of ComfyUI-Creepy_nodes:

• ComfyUI Manager: search "Creepy_nodes" and install.
• Manual:

  cd ComfyUI/custom_nodes
  git clone https://github.com/Creepybits/ComfyUI-Creepy_nodes.git
  

Restart ComfyUI. No dependencies beyond what the Gemini node already needs.

Troubleshooting

Empty dropdown = evaluate_people.json didn't load; ERROR: Prompt file not found. in the output = you renamed or deleted a persona file. Both fixed by restoring files from the repo. And if a persona run comes back as an error string instead of a verdict, that's the safety filter - loosen safety_threshold on the Gemini node or pick the _nice persona.