Text Strength Multiplier

Hand-editing weights in a long prompt is the ComfyUI equivalent of changing a dozen settings by hand. Text Strength Multiplier, from the illumorae pack (CorvaeOboro's text tools), does it in one shot: it takes your prompt, multiplies every parenthesized weight - (cat:1.3) → (cat:1.5) - while leaving <lora:...> tags untouched, and it can enforce individual caps, minimums, and a total weight budget on top. One number in, whole prompt rescaled.

The LoRA protection is the part worth paying attention to. Lora tags live inside angle brackets, not parens, and the node temporarily swaps them for placeholders before it touches any weight, so your <lora:style:0.8> strengths never get caught in the multiplier sweep. That's the difference between "scale my emphasis" and "silently break my loras."

How it works

The processing pipeline, from the source: find every weighted group (the (word:strength) pattern) and rewrite each strength by the multiplier. Then, in order:

• individual_cap_enabled / individual_cap - clamp each section's new strength to a max (default 1, off).
• individual_min_enabled / individual_min - floor each section at a minimum (default 0, off).
• total_cap_enabled / total_cap - sum all weights, and if the total exceeds the cap, rescale the whole prompt proportionally to fit. Minimums are hard floors; if the floors alone exceed the cap, everything goes to its minimum.

Caps apply before minimums, so a config where cap < min resolves to the minimum for every section - the code says so explicitly, and it's the kind of edge case you only discover after confusing yourself once.

Inputs and outputs

• text - the prompt (multiline).
• multiplier - the single knob (default 1, range 0–100). Above 1 amplifies, below 1 softens.
• The four toggle+value pairs above.

Output: modified_text, a plain STRING you wire into your CLIP text encode. There's no section_info here - check the result by eye or by feeding it to the pack's Text Show Multiline Color.

Installing it

ComfyUI Manager → search "illumorae" → install, or:

cd ComfyUI/custom_nodes
git clone https://github.com/CorvaeOboro/ComfyUI_illumorae

Restart. No models, no downloads.

The honest caveat

Weight syntax (word:1.3) is a SDXL-lineage thing. On the LLM-encoded models of 2026 (Flux, Z-Image, Anima), the encoder wrapper passes disable_weights=True and those weights are silently discarded - so scaling them changes nothing. If your checkpoint is Illustrious, NoobAI, Pony, or any SDXL finetune, this node earns its keep for batch weight sweeps. If you're on Flux, it's a no-op and you should skip it. Also note it only rescales weights that already exist; plain words without an explicit :strength don't get new weights unless the pack's sibling variant tooling does the wrapping for you. Nice tool, right model family required.