Text Parse Noodle Soup Prompts

"Noodle Soup Prompts" (NSP) sounds like a joke name, and it kind of is, but it's a real, specific thing: a curated JSON word bank the pack's author, WASasquatch, built years back with a collaborator (MisterRuffian) back in the Disco Diffusion era, before Stable Diffusion's wildcard culture had really settled on a standard. Text Parse Noodle Soup Prompts scans your input text for NSP tokens and swaps them for a word pulled from the matching category in that pantry - per the README, it "parses NSP in a text input." Think of it as a wildcard system that predates most of what people now call wildcards.

Where this actually fits

WAS Node Suite has an NSP-aware CLIP encode node too - CLIPTextEncode (NSP) - that does the same parsing but bakes it directly into text conditioning. This node is the standalone version: it just returns the parsed text, so you can inspect it, save it, concatenate it with something else, or run it through a second transform before it ever reaches a sampler. If all you need is "parse the tokens and encode," the combined node is fewer wires; if you want to see or manipulate the expanded text first, this is the one to use.

It's easy to confuse NSP with the more familiar __wildcard__ style from Automatic1111, and WAS's own nodes actually support both - they're just different systems living in the same node family. The author clarified this directly on Reddit when someone asked what NSP even was: switch the relevant mode to "Wildcards" for standard A1111-style folder wildcards (pointed at via wildcards_path in was_suite_config.json); NSP is the separate, bundled pantry with its own fixed vocabulary. Mixing up the two syntaxes is the single most common source of "why isn't this doing anything" confusion with this node.

Worth being honest about relevance: NSP is a genuinely old, niche corner even inside a pack that's mostly niche now. Discussion of it online is thin, and tools like the Dynamic Prompts extension have largely taken over the "randomize part of my prompt" job for most people. If you're already invested in A1111-style wildcard files, you may not need this specific node at all - but if you've got an old workflow built around NSP tokens, this is what parses them.

How it works

The node reads the input text, looks for NSP terminology, and for each match pulls a word from the corresponding category in the bundled pantry, seeded so the same seed reproduces the same expansion. No network calls, no model - it's a lookup against JSON data shipped inside the pack.

The inputs and outputs that matter

Text containing NSP tokens goes in; parsed text comes out as a string, ready to feed into a CLIP encode, a concatenate node, or wherever else expanded prompt text belongs. WAS's NSP-aware nodes are seed-driven for reproducibility, in line with how the rest of the pack's randomized text nodes behave.

How to install it

Ships with WAS Node Suite - the NSP pantry itself is bundled JSON, no separate download needed.

ComfyUI Manager: search "WAS Node Suite," install, restart. Manually:

cd ComfyUI/custom_nodes
git clone https://github.com/WASasquatch/was-node-suite-comfyui

then install requirements.txt against your ComfyUI Python and restart.

Common issues & troubleshooting

Nothing gets expanded. Almost always a syntax mismatch - you're either using __wildcard__ syntax against the NSP parser (or vice versa), and the two systems don't read each other's tokens. Confirm which mode/node you're actually using before assuming the pantry is broken.

Whole pack won't import. The pack-wide WAS issue: unmaintained since December 2023, and a ComfyUI update can throw "Import Failed" for every WAS node from a dependency clash. Reinstall requirements.txt against the correct Python environment and restart.