Weighted Prompt
Stop hand-typing (tag:1.3) and just slide a number
- STRING
Everyone who's tuned a CLIP-era prompt knows the drill: you hand-write (flower crown:1.25), (delicate smile:0.8), masterpiece into a text box and pray your parentheses balance. Weighted Prompt exists so you don't. You get three prompt slots, each with its own weight slider, and it assembles the properly-formatted string for you.
Here's the mechanism, straight from the source. Each prompt_N is trimmed (this pack's standard whitespace/commas cleanup), then compared against its weight_N. If the weight is 1.0, the prompt is emitted bare. If it's anything else, it's wrapped as (prompt:weight), formatted to three decimals. The three results are joined with ", ", and there's a multiplier (0–2) that scales every weight at once - so you can write your three concepts, then push all of them up or down together without touching each slider. A prompt with weight 0.0 or an empty prompt gets skipped entirely.
The node's own comment contains the one piece of ComfyUI lore you should internalize: an explicit weight is final and ignores surrounding brackets. ((cat:1.25), dog, bird:0.5) evaluates to cat:1.25, dog:0.5, bird:0.5 - the inner explicit weight wins over the nesting, so stacking parentheses around an explicit weight does nothing. Good to know before you start piling on.
The inputs that matter
prompt_1/prompt_2/prompt_3- your text blocks (multi-line, optional)weight_1/weight_2/weight_3- 0 to 2, step 0.1; 1.0 = no wrappermultiplier- 0 to 2, scales all three weights at once
One output: STRING, ready to feed a CLIP Text Encode.
Where it works - and where it's dead weight
(tag:1.3) weighting is a CLIP-era feature. It works on SD 1.5, SDXL, Illustrious, Pony - the models whose text encoders were built around the syntax. It does not work on Flux or any LLM-encoded model, where the parentheses and colons are passed to the encoder as literal punctuation and actively make your prompt worse. If you're on Flux, don't reach for this node; say what you want in words instead.
Installing it
Same as the rest of the pack - no dependencies beyond ComfyUI's own, no models. ComfyUI Manager → search ComfyUI-Prompt-Helper, or git clone https://github.com/elypha/ComfyUI-Prompt-Helper into custom_nodes and restart. It shows up under prompt_helper.
Honest verdict: for the handful of models it targets, it's nicer than hand-typing weights - the multiplier alone is worth it. For Flux users it's a polite no-op with a warning sign on it.
Inputs (7)
| Name | Type | Default | Description |
|---|---|---|---|
| prompt_1opt | STRING | — | |
| weight_1opt | FLOAT | 1.00–2 | — |
| prompt_2opt | STRING | — | |
| weight_2opt | FLOAT | 1.00–2 | — |
| prompt_3opt | STRING | — | |
| weight_3opt | FLOAT | 1.00–2 | — |
| multiplieropt | FLOAT | 1.00–2 | — |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| STRING | STRING | — |