Prompt Styler (Arctenox's Essentials)

There are a dozen "prompt styler" nodes floating around the ecosystem, and they all do the same thing: pick a vibe, get a chunk of tags pasted into your prompt so you don't have to remember them. This is Arctenox's version, and it's a decent one - 52 presets covering photography (cinematic, studio portrait, macro, HDR), anime and illustration (anime, manga, chibi, webtoon), traditional art (oil painting, watercolor, sketch), digital art, sci-fi and fantasy, vintage eras, and special effects like glitch and neon.

Key thing to understand before you install it: this is a STRING node. It outputs text - positive_prompt and negative_prompt - not conditioning. It sits before your CLIP Text Encode node, which is exactly where a style node belongs. You can also use it to build a prompt you're going to paste elsewhere, because the output is just plain text.

How it works

The mechanism is simple string assembly. It builds a positive from your base_prompt, plus a quality tier's tag block, plus the style's positive template, plus anything in custom_positive - all joined with commas. On the negative side it takes your base_negative and appends the style's negative suggestions, but only if use_negative_suggestions is on. Then it resolves wildcards across the entire final strings, using wildcard_seed for reproducibility (−1 = a fresh pick every run).

Two details worth knowing:

• The quality tiers are tag blocks with real teeth: Draft → "good quality" up to Ultimate → "masterpiece, best quality, ultra detailed, 8k uhd, professional, award winning, trending on artstation."
• Wildcards work everywhere in the node - {cat|dog} inline picks and __filename__ file tokens - and they're resolved on the finished string, so styles can even contain wildcards if you edit the source.

The inputs that matter

• base_prompt - your subject. This is the field that matters most; everything else wraps around it.
• base_negative - pre-filled with the giant inherited boilerplate block (bad anatomy, bad hands, watermark, the works). It's editable, and you should feel free to trim it.
• style - the dropdown with all 52 presets.
• quality_tier - None / Draft / Standard / High / Ultimate.
• use_negative_suggestions - toggle the style-appropriate negative tags.
• custom_positive / custom_negative (optional) - add your own bits on top of the preset.

The outputs

positive_prompt and negative_prompt, both plain STRINGs. Wire them into a CLIP Text Encode node and you're sampling.

Install

ComfyUI Manager, search Arctenox's Essentials, install, restart. Or:

cd ComfyUI/custom_nodes/
git clone https://github.com/Arctenox/Arctenoxs-Essentials_ComfyUI

Restart. No model downloads, just torch, numpy and optional psutil.

Gotchas

• Presets are tag blocks, not magic. They're tuned for SDXL-lineage models where real CFG is running. On guidance-distilled models that run at CFG 1, the negative side of any style does nothing at all - that pass isn't even computed. If your checkpoint is a Turbo/Lightning/distilled release, restate things as positive tags and don't worry about the negatives.
• The default negative is the inherited forty-keyword block the community has been trimming for years. It's harmless on SDXL, inert on CFG-1 models - but either way, test whether yours actually earns its place before keeping it.
• If a style "isn't working," remember it's just a string. Open the output and read what it actually built - you'll usually see the problem immediately.