Russia Attire

There's a specific failure mode where you ask for a "traditional Russian outfit" and the model hands you a generic medieval dress with a fur trim. The Russia Attire node from ComfyUI Globetrotter Nodes closes that gap the only honest way: by naming the actual garments - Sarafan, Kosovorotka, Kokoshnik, Ushanka, Valenki, Lapti, Poyas - and letting their descriptions carry the visual detail instead of hoping the model fills it in.

What it is

A character-prompt generator, one of the country nodes in the manifestations/comfyui-globetrotter pack. Pick a person and a wardrobe, get one prose prompt STRING out, feed it into CLIP Text Encode as your positive prompt.

How it works

The pack's standard engine: the node is generated at load time from JSON files in data/, it assembles the prompt in priority order (subject → appearance → emotion → attire → pose → setting → atmosphere → composition → quality), it gender-filters the attire dropdowns, and it expands every garment name into a description. Russia's data is thin like several others in the pack - no regional appearance data, no country poses or landmarks, four attire categories - so treat it as an attire-and-subject generator rather than a scene builder.

The inputs that matter

• Upper_Body - Sarafan (the sleeveless jumper-dress), Kosovorotka (the collarless shirt), Telogreika (the padded jacket). The sarafan is the one most people are after; models routinely butcher it without the description.
• Head - Kokoshnik (the traditional headdress), Ushanka, Papakha.
• Feet - Valenki (felt boots), Lapti (bast shoes). Waist - Poyas (the sash).
• gender and age set the subject and filter which garments appear - the kokoshnik and sarafan are female-tagged, for instance.
• randomization (off/light/moderate/full) and seed - off pins results, full ignores the seed for maximum variety.
• lora_trigger, custom_prompt, and experimental_llm_rewrite round it out; the last needs transformers and downloads distilgpt2 on first use.

Output

One prompt STRING → CLIP Text Encode (Prompt) → KSampler. Standard wiring.

Install

cd /path/to/ComfyUI/custom_nodes
git clone https://github.com/manifestations/comfyui-globetrotter
cd comfyui-globetrotter
pip install -r requirements.txt   # optional - only for LLM rewrite

Restart ComfyUI, or search "Globetrotter" in ComfyUI Manager. No model downloads - the Russian data ships as JSON in the repo.

Troubleshooting

• Only four attire categories: correct - this node covers Head, Upper_Body, Feet, and Waist. If you need more, the pack is built to be extended: add items to data/attire/*/ru.json and restart.
• Ushanka renders comically huge: it's a strongly archetyped item in base models; pair the description with a tighter composition (portrait or close-up) rather than a wide shot.
• Output varies between runs: randomization defaults to moderate; set it to "off" and fix the seed for reproducibility.
• experimental_llm_rewrite silently does nothing without transformers installed.

Narrow job, done properly: authentic Russian traditional garments described well enough that your encoder stops guessing.