Simple Prompt (full-pipe)

Simple Prompt (full-pipe) is what it says: the mudknight prompt node with the training wheels off and the magic removed. It takes a FULL_PIPE, conditions positive and negative text, strips comment lines, handles <lora:...> syntax - and then stops. No quality-tag presets, no style switcher, no character matcher. That's deliberate: the author made it for Anima, whose prompt order is too specific for the preset machinery to help with, and the docs are blunt that detailing and preset-tagging generally make Anima results worse.

The one feature that makes it worth a second look is that it remembers your prompt. It writes the raw positive and negative text into the pipe's meta, and the Save (Full Pipe) node reads that to embed A1111-compatible metadata in the output image. If you're the kind of person who drags a finished PNG into A1111 later and expects the prompt to be there, that's the reason to pick this over a plain CLIPTextEncode. A1111 metadata isn't standardized in ComfyUI, so the author built the whole save path around it - this node feeds that path.

What you set

• positive / negative - multiline text, same as the big prompt node. #-prefixed lines are dropped; <lora:name:strength> gets stripped from the text, parsed, and applied to the model/CLIP in the pipe. A third colon gives separate model:CLIP strengths.
• trigger_words - optional force-input string appended to the end of the positive text. Wire a character or style node into it.
• negative_to_negpip - the one interesting toggle. Turn it on and the negative box's tags get inverted and folded into the positive prompt as negpip syntax (something like (tag,:-1)), leaving the actual negative prompt empty. This is the negpip trick for models where the negative prompt is weak or where you want negative presence on a distilled/low-CFG model. It requires the CLIP NegPip module from pamparamm/ComfyUI-ppm to actually do anything.

How it works

It strips comments and newlines, appends trigger words, extracts LoRA syntax, encodes positive and negative with the built-in CLIPTextEncode, applies the LoRAs to the model, and returns an updated full_pipe with positive, negative, model, clip, and meta (with positive_text and negative_text) all set. The text metadata is captured before LoRA extraction, so the A1111 string you get back matches what you typed, not the cleaned version - a small touch that matters if you've ever opened a workflow image and found your <lora:...> tags gone.

Output is one full_pipe. Wire it into Base (full-pipe) or Upscale (full-pipe).

Installing

Same pack as everything here: ComfyUI Manager → search comfyui-mudknight-utils, or

cd ComfyUI/custom_nodes
git clone https://github.com/mudknight/comfyui-mudknight-utils

Restart. The negative_to_negpip toggle needs ComfyUI-ppm from pamparamm - the node itself doesn't require it, but the toggle is a no-op without it, so install ppm if you want that behavior. No model downloads.

Gotchas

• Toggle does nothing → ppm isn't installed. The node won't tell you; it just proceeds as if negpip worked.
• Negative prompt dead at low CFG - not a bug in the node. On distilled/CFG-1 models a negative prompt is mathematically inert, which is exactly the case negpip exists for.
• Don't expect quality-tag injection - that's the other prompt node's job. If you switch here from Prompt (full-pipe) and your score_9 tags vanish, you turned off the wrong feature.

Zero impressions on comfy.icu, tiny pack, but if you run Anima or you care about your prompt text surviving into PNG metadata, this is a genuinely pleasant little node.