Powder Styler

The Powder Styler is the E2GO "Powder" pack's answer to SDXL prompt-styler nodes: a library of styles you can stack, toggle, and feed straight into Powder Conditioner as JSON. If you've used the classic prompt-styler packs, you know the model - a style is a prefix, a suffix, and a negative that wrap your prompt. What this one does differently is treat styles as data, not as a giant dropdown of pre-baked combos.

It ships with a starter library (including SDXL prompt-styler-format JSON files), but the real power is that your own styles are just .json files in a folder. Add one, wait two seconds, and it's in the dropdown - no restart, no code.

How it works

Each style in the library is a small object with four fields: a name, a prefix, a suffix, and a negative. On the node you build a list of up to 20 style slots; each slot has an Enabled toggle, a style picker, and per-slot Use Positive / Use Negative toggles. The node merges all enabled styles, deduplicates tags, and emits a single JSON blob describing the combined prefix, suffix, negative, and position - plus a plain-text style_text output.

The style_position combo controls how the style wraps your prompt:

• Wrap prompt - prefix + prompt + suffix (the classic).
• Before prompt - prefix + suffix + prompt.
• After prompt - prompt + prefix + suffix.

Position lives in the emitted JSON, so the Conditioner knows how to assemble the final text.

Adding your own styles

Drop a JSON file into e2go_nodes/styles/. The filename doesn't matter - any .json in that folder is picked up within ~2 seconds, no restart. The format:

[
  {
    "name": "My Cool Style",
    "prefix": "text before prompt",
    "suffix": "text after prompt",
    "negative": "added to negative prompt"
  }
]

The old SDXL prompt-styler format is also supported, where prompt contains the whole wrapper with a {prompt} placeholder and negative_prompt is the negative. And since both Styler and Powder Grid Saver share the same styles cache, a new style is visible to the grid labels at the same time it appears here.

Inputs and outputs

• style_position - wrap / before / after.
• use_positive / use_negative - master switches for whether styles contribute to positive and negative text at all. Per-slot toggles override these.
• style (output) - JSON with prefix, suffix, negative, position. Wire this into Powder Conditioner's style input.
• style_text (output) - the combined prefix+suffix as plain text, handy if you want to inspect or use it without the conditioner.

Installation

One install, all eight nodes:

cd ComfyUI/custom_nodes
git clone https://github.com/E2GO/e2go-comfyui-nodes.git e2go_nodes

Restart ComfyUI, or search e2go_nodes / E2GO in ComfyUI Manager. No extra Python dependencies, verified against ComfyUI 0.17+.

Common issues

The usual stumble is expecting the style JSON to be human-readable. It isn't - style is machine food for the Conditioner, and if you want to see what a style actually does to your text, read style_text or the Conditioner's final_positive output instead.

Second, position is a per-node setting, not per-style. If you mix styles that need different positions, the node can't express that - pick one position for the whole slot list and design your styles around it.

And a real-world prompt-engineering note: style suffixes that pile on quality tags matter differently depending on your model. On SDXL-lineage models (Illustrious, Pony, NoobAI) the tag-style suffix is doing real work; on newer LLM-encoded models much of that vocabulary goes inert. The Styler doesn't care - it just assembles text - but if your styles stop having an effect on a new checkpoint, the style text isn't the problem, the encoder is. The dedup it does on tags keeps your prompt from bloating with repeated quality tags, which helps on both.