🙈 FoW - Content Weight

One slider for the "what's in frame" negatives

The FoW Suite's nine categories split a prompt into labeled buckets, and Content is the "what's actually in the picture" bucket: nsfw, suggestive, multiple characters, photorealistic-when-you-don't-want-it, that kind of thing. FoW - Content Weight is a single dial that scales how hard that whole bucket pushes, without you re-encoding anything.

It's part of FoW_Suite_LIGHT, the free tier of the "Force of Will" suite by SirWillance - a self-taught dev who built the pack as his first project over two months and licenses it MIT. The paid Standard/Pro tiers swap in "more inputs and sliders," which is a polite way of saying Light tier keeps this as basic as possible. For a beginner that's not a bad thing.

How it works

Mechanically it's dead simple, and it's worth knowing because it tells you what the dial actually does. The node takes a CONDITIONING tensor and multiplies every entry by your weight:

weighted.append((tensor.clone() * Weight, metadata))

The whole Content slice gets scaled as one block - there's no token-level weighting in Light tier. A Content Category / ContentVillain node produces the conditioning; this node just decides how loudly it speaks.

The inputs that matter

Two inputs, one you'll actually set:

• RAW Condition (CONDITIONING) - the Content conditioning to scale. Plug in the Content Conditioning output from a Content Villain or Content Category node.
• Weight (FLOAT, 0–2, default 1, step 0.1) - the dial. 1.0 = no change.

The one output, W8ed Content Condition (CONDITIONING), feeds a Fusion node or straight into a KSampler's negative input.

Installing it

The suite installs as one pack:

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

Restart ComfyUI and the nodes appear under "🧿 FoW - Light". ComfyUI Manager also has it - search "Force of Will Suite Light". There are no extra Python dependencies and no model downloads; everything runs off the CLIP model your checkpoint already loads.

Where people get burned

Same traps as every weight node in this suite. Cranking to 2.0 on a safety-flavored bucket like Content tends to over-correct and flatten the image - the model over-applies and you lose more than you wanted. Nudge, don't slam. And be realistic about the checkpoint you're on: at CFG 1 on a distilled/turbo model, negative conditioning is basically inert, so a Content weight of 1.8 will do close to nothing. On normal SDXL-lineage checkpoints at CFG 4–9, though, it works exactly as labeled. It's a fine, honest little knob - just don't expect it to rescue a negative prompt that's already doing nothing.