2🐕Character category

If you don't have the Danbooru tag wiki memorized, describing a character properly is a chore - you know you want "twintails" and "hair ornament" but you're bouncing between browser tabs trying to remember the exact tag spelling. EG_TSCDS_RW ("Character category") turns that whole lookup into fourteen dropdowns you click through instead.

How it works

It's one node that bundles fourteen head-to-toe categories: Type (broad archetype - 1girl, solo, chibi, loli, maid…), Character in the work (named Danbooru characters - Alice Margatroid, Akemi Homura, Aerith Gainsborough…), Face, Rare hairstyle, Rare hairstyle man, Upper body decoration, Lower body decoration, Full body decoration, Hair, Head accessories, Ears, Neck, Clothing, and Shoes and socks. Every category defaults to "None" and carries its own weight slider, so you only pay for what you actually pick. Whatever you select gets assembled into one weighted prompt string on the output - this is the Danbooru-style (tag:weight) syntax that SD 1.5, SDXL, Illustrious, Pony and NoobAI checkpoints all read natively.

The inputs and outputs that matter

You'll rarely touch more than a handful of the fourteen dropdowns per generation - that's the point, it's a menu, not a form. Each category pairs with a *weight slider (0.1–2.0, default 1); the KB's practical range for tag emphasis is 0.5–1.5, since below 0.5 a tag barely registers and above 1.5 you start getting oversaturation, so don't crank every slider to max just because it's there. There's a global random toggle (default no) - flip it to yes and the node ignores your manual picks and rolls a random selection across all fourteen categories instead, with seed making that reproducible. The single output, prompt (STRING), is a ready-to-concatenate fragment - wire it into a text-concat node alongside your other prompt pieces, or straight into a CLIP Text Encode if it's all you need.

How to install it

Search Comfyui-ergouzi-Nodes in ComfyUI Manager, or do it by hand:

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

Restart ComfyUI. One thing worth knowing before you lean on this pack: the author's own README says this English-named repo is frozen - "this library will no longer be updated" - because development moved to a Chinese-named sibling repo (Comfyui-ergouzi-DGNJD) that has more nodes. What comfy.icu ships is this frozen snapshot, not the actively maintained one.

Common issues & troubleshooting

The weight syntax does nothing on newer models. (tag:1.3)-style emphasis is a CLIP-era convention. On an LLM-encoded checkpoint - Z-Image, Flux 2 Klein, Anima, Krea 2, ERNIE - that syntax isn't parsed as a weight at all, it's passed through as literal text sitting in your prompt, which is arguably worse than leaving the weight off. If you're on one of those models, treat the weight sliders as decoration and consider dropping them to 1 across the board, or just describe the character in plain language instead of reaching for this node.

Since it's frozen, don't expect fixes. A no-longer-maintained single-file node pack is fine until ComfyUI's frontend changes underneath it - the Nodes 2.0 rewrite in late 2025 broke UI widgets in packs that were actively maintained, and this one isn't. If a category dropdown stops rendering after a ComfyUI update, that's the likely cause, and there's no upstream fix coming for this repo.

Output feels diluted with everything filled in. Stacking all fourteen categories at weight 1 gives the model fourteen competing instructions. Pick the three or four that actually matter for this image and leave the rest on "None" - the node's whole value is letting you be selective without hunting for tag spellings, not filling every slot because you can.