Anima Prompt (full-pipe)

If you've ever stared at a wall of danbooru tags trying to remember which line was the character and which was the style, this node is for you. Anima Prompt (full-pipe) turns one giant prompt box into five labeled fields - num_characters, characters, copyrights, style, and prompt - splices them into a single positive prompt, strips out lines you've commented with #, and applies any <lora:file:strength> syntax it finds. It's built for the anime model families (Illustrious, Pony, NoobAI, and the newer Anima/Qwen line) where prompts are tag soup and LoRAs are how you control identity.

It takes a full_pipe, reads the model and CLIP out of it, and returns the same pipe with fresh positive and negative conditioning plus updated model/CLIP if a LoRA was in the text. The categories concatenate in a fixed order (characters, then copyrights, then style, then the main prompt), which gives you a stable structure for your tags.

The inputs that matter

• The five positive fields - all optional multiline strings. Use # at the start of a line to comment it out without deleting it (escape it as \# if you genuinely need a # in the prompt). The pack's example Anima workflow uses exactly this: # Quality tags as a comment line, then the real tags.
• negative - the negative prompt, also comment-aware.
• negative_to_negpip - a toggle that converts your negative prompt into negative-weight positive syntax, which requires the CLIP NegPip module from ComfyUI-ppm. It's a real feature, not a gimmick: it lets you say "not blue hair" instead of hoping the negative encoder understands. Skip it if you don't have ppm installed - the node works fine with it off.
• <lora:file:strength> - drop these anywhere in a field and the node extracts them, loads the LoRA, and patches the pipe's model and CLIP. This is how the pack's Anima workflow loads its RDBT stabilization LoRA at 0.5 strength without a separate LoRA node.

Output is the updated full_pipe - wire it to the pack's generation node and you're done.

Installing it

It's in mudknight-utils - ComfyUI Manager (search comfyui-mudknight-utils) or:

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

The only optional dependency is ComfyUI-ppm if you want negative_to_negpip. Everything else is core.

Where people get burned

The most common mistake is expecting the node to do something to a model that isn't in the pipe - it needs a real pipe from a loader, not just any CLIP. And if your negative tags aren't landing, remember the anime-model quality-tag convention (the score_9 style tags for Pony/Illustrious) lives in your prompt fields, not in this node - it just splices what you type. One more: the negpip toggle rewrites your negative into the positive; if results look over-constrained, drop the toggle rather than fighting the syntax.