Analyze Prompt

Here's a workflow-automation trick that's genuinely fun: put ☹ in your prompt when you want a face detailer to run, and ✌ when you want a hand refiner - and let a node translate those emoji into actual workflow decisions. That's the entire personality of Analyze Prompt. It takes a prompt, looks for those two characters, strips them out of the text, and sets a pair of booleans the rest of your graph can branch on.

Input: one required prompt (STRING). Outputs:

• prompt - the cleaned text, emoji removed.
• facedetailer_enabled - true if the prompt contained ☹.
• handrefiner_enabled - true if the prompt contained ✌.

The mechanism is exactly what the README says, no more: a literal string check. ☹ in the prompt → facedetailer flag on and the character removed; ✌ → handrefiner flag on and removed. Then a second, separate rule kicks in for NSFW cues: if the prompt contains certain explicit terms (nude, sex, masturbation, and friends), the node appends , explicit and , uncensored unless they're already present. Wire the two booleans into a switch or gated branch and you've got prompt-driven detailing: type ☹, the face detailer runs; leave it out, it doesn't.

Where this shines and where it's just silly:

• The good use: batch prompting with a character that's a cheap, portable instruction. You're generating 200 prompts; you want faces fixed on the portrait-ish ones and not on the wide shots. Encoding that intent as a character in the prompt is the kind of thing that makes a batch workflow feel alive. And the NSFW auto-tagging is aimed at uncensored-model prompt hygiene - get the explicit markers in before the model's own heuristics guess wrong.
• The reality check: the emoji are a private convention of this one pack, and the matching is blunt. ☹ sets the face flag no matter the context, and the NSFW detection is a fixed list of English terms matched as whole tags. It won't catch paraphrases, it won't catch Japanese prompts, and it appends explicit, uncensored even if you're on a model that doesn't want them. Treat the NSFW rule as a habit, not a guarantee.

It's part of 2daadv's ComfyUI-GadgetNodes Gadget/prompt family (MIT, one developer, brand-new pack with no community reputation). Install via ComfyUI Manager (search "GadgetNodes") or clone + pip install -r requirements.txt + restart. It's pure Python with no frontend extension, so it runs identically on the legacy canvas and Nodes 2.0 - the flags are just outputs you wire, not UI magic.

One warning that applies to the whole emoji approach: the flags only do something if the downstream graph actually uses them. This node emits decisions; you still have to connect facedetailer_enabled to a switch that gates an Impact-Pack-style detailer. Skip that wiring and you've got a node that removes an emoji from your prompt and nothing else.