Gothic Styler (Advanced)

Building a spooky scene and tired of hand-typing "dark cathedral, moody candlelight, chiaroscuro" into every prompt? GothicStylerAdvanced is a single dropdown of 91 gothic presets - Dark Cathedral, Chiaroscuro Masterpiece, Gothic Tarot Cards, Haunted Mansion, Phantom Opera, Victorian Mourning - that wraps your prompt in gothic language and appends a matching negative. It's the Advanced variant from aegis72/comfyui-styles-all, the preservation clone of azazeal04's removed SDXL Prompt Styler.

What "Advanced" actually means here

The plain Gothic Styler gives you one positive box and two outputs. This one splits the positive into G (global) and L (local) strings and hands you six outputs - the idea being that SDXL has two text encoders (OpenCLIP-G for the "big picture" text, CLIP-L for detail), and you feed each one its own string. In this pack's style files, the G/L split is driven by templates that contain a . separator, and there's a negative_prompt_to dropdown so you can send the negative to Both encoders, G only, or L only.

Real talk about these particular templates: the gothic styles are single-part wrappers (something like (Gothic-style chiaroscuro masterpiece) {prompt}), so there's no . split happening. The Advanced node mostly adds plumbing, not magic.

A gotcha the docs won't tell you

I ran the actual code, and the two positive boxes trade places. Whatever you type into text_positive_l is what gets styled and exits through the text_positive_g output; your text_positive_g input flows through unstyled to the text_positive_l output. The combined text_positive output still contains both, joined by ..

Practical upshot:

• Simplest path: put your whole prompt in text_positive_l and use the combined text_positive output into a normal CLIPTextEncode.
• If you're on SDXL and want the true two-encoder split, use the combined text_positive plus text_negative into CLIPTextEncodeSDXL's text_g/text_l, or feed text_positive_g to the G encoder and text_positive_l to the L encoder as labeled.
• If all you need is one dropdown on one box, honestly the non-Advanced Gothic Styler is the less confusing pick.

Inputs and outputs

Inputs: text_positive_g, text_positive_l, text_negative, the Gothic dropdown, negative_prompt_to (default Both), and log_prompt (default on - prints before/after to the console). Outputs: text_positive_g, text_positive_l, text_positive, text_negative_g, text_negative_l, text_negative.

Installing it

ComfyUI Manager (search ComfyUI-styles-all), or:

cd ComfyUI/custom_nodes
git clone https://github.com/aegis72/comfyui-styles-all

Then restart. No requirements.txt, no models - just Python stdlib and JSON.

Where people get burned

• The G/L swap above - if your styled output looks like it's missing the subject, that's why.
• Restart after editing JSON. The dropdown is rebuilt from data/Gothic/Gothic.json at load time. You can edit it or add your own file in that folder to grow the menu.
• Flux shrugs at it. These keywords were tuned for SDXL's CLIP encoders; on LLM-encoded models the gothic styling lands much weaker.