2🐕Style category

"Give me this in a specific illustrator's style, rendered like a graphic novel, with a bit of motion blur" is three or four separate tags most people don't have memorized. EG_TSCDS_FG ("Style category") is the ergouzi pack's answer for the "what does this look like" half of a prompt, as opposed to the "what's in it" half covered by the character and item pickers.

How it works

Eight categories, one node, same pattern as the rest of the "Fixed" family: Graphic effects (blurry, sparkle, motion lines, lens flare, jpeg artifacts…), Art style (sketch, cosplay, traditional media, watercolor, realistic…), Theme (Hyper Real, Photorealism, Surrealism, Daydreampunk…), Art unconventional (alternate costume, genderswap, humanization, aged up…), Illustration style (Anatomical Drawing, Caricature, Crosshatch, Graffiti…), Artist (Zhang Daqian, Frida Kahlo, David Hockney, William Eggleston…), Film director (mostly architects here, oddly labeled - Zaha Hadid, Rem Koolhaas, Antoni Gaudi…), and Coding method, which is actually the Danbooru censorship-tag vocabulary (censored, mosaic censoring, bar censor, uncensored, hair censor…). Each category has its own weight slider and defaults to "None." Everything you pick combines into one weighted prompt string.

The inputs and outputs that matter

Realistically you'll pick one or two style anchors - an Art style plus maybe an Artist, or a Theme plus a Graphic effect - rather than filling all eight. Weight sliders run 0.1–2.0 (default 1); tag-emphasis testing across the community puts the useful range at roughly 0.5–1.5, so treat anything past that as diminishing returns rather than a stronger effect. random (default no) plus seed let the node roll random picks across all eight categories instead of your manual choices, handy for exploring style combinations you wouldn't have thought to try. Output is a single prompt (STRING) to concatenate with the rest of your prompt or feed directly to a CLIP Text Encode.

How to install it

ComfyUI Manager: search Comfyui-ergouzi-Nodes. Manual install:

cd ComfyUI/custom_nodes
git clone https://github.com/11dogzi/Comfyui-ergouzi-Nodes.git

Restart ComfyUI. One caveat from the author's own README: this English-titled repo is no longer updated, with active development continuing under a Chinese-named sibling pack. What you're installing is the frozen version.

Common issues & troubleshooting

Weighted style tags do nothing on LLM-encoded models. (watercolor:1.4) only functions as emphasis on CLIP-based checkpoints - SD 1.5, SDXL, Illustrious, Pony, NoobAI. On Z-Image, Flux 2, Anima, or Krea 2, the parentheses-and-colon syntax passes through as literal text instead of being interpreted as a weight, so a style prompt that reads fine on an older model can come out garbled on a newer one.

Named artists and directors aren't guaranteed to be represented well. Style transfer from a named-artist tag depends entirely on how well that name is represented in your checkpoint's training data - a base SDXL model and an anime-tuned Illustrious finetune won't render "Frida Kahlo" the same way, and some names may barely register at all. Treat the Artist and Film director lists as suggestions to experiment with, not guarantees.

Coding method feels out of place if you weren't expecting it. That category is standard Danbooru censorship vocabulary, not an image-quality setting - it's there because the underlying tag database includes it, not because it's central to what this node does. Leave it on "None" unless you specifically need it.