Nodes/FoW - Light/👤 FoW - Subject Category Light
ComfyUI Node

👤 FoW - Subject Category Light

The FoW Node That Actually Lets You Write — Character Builder for Prompts

By SirWillance·Created 2 years ago·Updated about a year ago· 5
👤 FoW - Subject Category Light
  • clip
  • Subject Conditioning
  • Subject Prompt
subject_base
subject_characteristics
subject_attire_and_gear
subject_details

Here's the FoW suite node I'd actually reach for. The Style Agent and the villains are all catalogue pickers - you choose from what the author curated. FoWLSubjectCategory is different: it's a freeform form for defining your character or subject. Four labeled text boxes, you type whatever you want, and it assembles them into one conditioning. No menu, no hidden vocabulary, no "tier" gating your options. It's the least gimmicky node in the pack.

What the four fields do

Hit "Open Subject Interface" and you get four multiline boxes. There's no strict schema - they're just a way to keep your thinking organized:

  • base - what the subject is. "a knight", "a tabby cat", "an astronaut".
  • characteristics - traits, build, expression, age.
  • attire_and_gear - clothes, armor, weapons, props.
  • details - the small stuff that separates a generic render from something specific.

You can fill one field or all four. Each non-empty field gets CLIP-encoded on its own and then the results are concatenated into a single conditioning - so the final prompt is one subject block, not four competing paragraphs.

How it works

Under the hood it's refreshingly plain. The node takes each filled field, strips any FoW / FoWs / FoWe separators you've typed (the author's delimiter trick for splitting prompts), encodes each piece with the standard CLIP tokenizer, then merges them with ComfyUI's ConditioningConcat. Empty fields are skipped entirely. Two outputs come out:

  • Subject Conditioning (CONDITIONING) - wire this into the KSampler positive, or into a Conditioning (Combine) node to stack it with the style/shot agents.
  • Subject Prompt (STRING) - the assembled, comma-joined text, if you want to see exactly what the model was told (or want to reuse it elsewhere).

The one required input is clip from your checkpoint loader, same as every other FoW category node.

Why this beats a plain text encoder

Mostly because of the workflow shape it forces. Splitting a subject into base / characteristics / attire / details is genuinely good prompting hygiene - earlier, more specific tokens get stronger attention in CLIP, and a well-structured subject block reads better to the encoder than a wall of comma soup. For the character-consistency workflows where you'll regenerate a character across poses and scenes, having the subject baked into one reusable conditioning block is legitimately convenient.

It also plays nicely with the pack's "modular" idea: encode your subject once with this node, style it with the Style Agent, and swap either without retyping the other.

Installing it

Standard FoW suite install, nothing special:

cd ComfyUI/custom_nodes
git clone https://github.com/SirWillance/FoW_Suite_LIGHT

Or ComfyUI Manager → "Force of Will Suite Light". Restart, done. No requirements file, no models.

Common gotchas

  • The text fields look like they don't exist. Like the rest of the suite, the four subject_* inputs are hidden behind the modal button. That's by design - but it also means you can't see or edit them without reopening the interface, and if the modal ever fails to render, the node looks like a brick.
  • The FoW separator is literal. The cleaning step only strips that exact string (and its FoWs/FoWe variants). If you type FoW expecting it to be preserved in the output text, it won't be - that's the point.
  • Empty-field behavior is fine, but not obvious. Leave everything blank and you get an empty conditioning, which is a silent no-op. If your image "ignores" the node, check that you actually filled something in.

One honest warning, though: this is a very young pack from a first-time coder with essentially no community footprint yet. The core idea is sound and the MIT license is generous, but don't expect the polish or the support threads of the big-name packs.

Category🧿 FoW - Light/📇 The 9 Categories/👍 Positive

Inputs (5)

NameTypeDefaultDescription
clipCLIP
subject_baseoptSTRING
subject_characteristicsoptSTRING
subject_attire_and_gearoptSTRING
subject_detailsoptSTRING

Outputs (2)

NameTypeDescription
Subject ConditioningCONDITIONING
Subject PromptSTRING