🎨 FoW - Style Weight

You've got a style conditioning from FoWLStyleAgent - but the style is fighting with your subject, or it's just a bit too loud. FoWLStyleWeight is a single slider that turns that whole style up or down. It's the volume knob for one conditioning block.

What it does

This is a "weight" node, so let's be precise about what that means here, because it's not what you might expect from Automatic1111. In A1111, (word:1.3) reweights a single token inside the prompt text. This node does something coarser: it takes a whole conditioning - the Style Conditioning output from the Style Agent - and multiplies the conditioning tensor by a factor. It's the entire style block getting louder or quieter, not individual words.

• RAW Condition - the conditioning you want to scale. Wire in Style Conditioning from a FoWLStyleAgent.
• Weight - a slider from 0 to 2, default 1, in 0.1 steps. 1 is pass-through, below 1 weakens the style, above 1 amplifies it.

The output is W8ed Style Condition (yes, "W8ed" - the author's shorthand for "weighted" is the actual output label), a CONDITIONING you feed into the KSampler or a Conditioning (Combine) node. Mechanically it's tensor.clone() * Weight per conditioning entry - a couple of lines in the pack's shared base class.

Where it earns its keep

The classic move in the FoW workflow is a subject block, a style block, and a shot block all fused together, and sometimes one dominates. Dialing Style Weight down to 0.7 while pushing Subject Weight up is a far more direct way to rebalance than rewriting prompts. Same story if a strong style like "art deco" keeps overwhelming everything else - clamp it with this instead of arguing with the text.

A couple of honest caveats:

• It's a sledgehammer, not a scalpel. The whole style gets scaled uniformly. If only part of the style is too strong, this won't help - you'd need per-token weighting inside the agent instead.
• Past ~1.3-1.5, conditioning scaling tends to oversaturate - colors blow out, details smear. Think of it as a gentle trim, not a way to force a 10x stronger style.
• One nice property: because it scales the conditioning tensor directly rather than rewriting prompt text, it works regardless of which text encoder your checkpoint uses. That's the opposite of text-weight syntax like (word:1.3), which modern LLM-based encoders simply discard.

Installing it

It ships with the rest of the FoW suite - no separate install:

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

or ComfyUI Manager → "Force of Will Suite Light", then restart. No dependencies beyond ComfyUI itself.

The usual traps

• Wiring the raw text instead of the conditioning. This node takes CONDITIONING, not STRING. If you try to feed it the Style Text output, ComfyUI will throw a type mismatch before you even run it.
• Expecting it to change the prompt text. It doesn't - the output is a reweighted conditioning, and the Style Text string is untouched by design.
• Same hidden-input pattern as the rest of the pack. The RAW Condition is right there on the node, but if you came from the agent nodes expecting a button, there isn't one; this is a plain utility node.

It's a small, boring, useful node - exactly the kind that disappears into a workflow and does its job. The pack around it is young and quirky, but this one's hard to break.