Chinese Converter (Simp⇄Trad)

If you work with workflows from both mainland-China and Taiwan/HK creators, you've hit the two-Chinese-scripts problem: a prompt written in Simplified characters, your model was trained with Traditional - or the other way around - and character-level details drift when you convert naively. ChineseConverter is a small node that does the conversion properly. It's the kind of tool that looks trivial until you've watched a bad converter mangle 「乾」 into 「干」 and realize script conversion is actually linguistic, not just a font swap.

How it works

The node wraps OpenCC (opencc-python-reimplemented), the standard open-source library for Simplified/Traditional conversion. OpenCC isn't a one-for-one character map; it applies phrase-level rules, so it correctly handles the words that change meaning between scripts rather than just swapping glyphs. A single boolean, simp_to_trad, picks the direction:

• simp_to_trad = true (default) - Simplified → Traditional.
• simp_to_trad = false - Traditional → Simplified.

The library is loaded lazily: the node auto-installs it on first use and then wants a restart so the package is importable (same first-run dance as the pack's other Chinese node). After that, conversion is local and instant - no API, no key, works offline.

Inputs and output

• input_text - multiline; paste the prompt, caption, or script.
• simp_to_trad - the direction switch.

Output: converted_text (STRING), ready to wire into a prompt input or a Show Text node. Multiline input means you can run a whole caption through at once.

Installing it

cd ComfyUI/custom_nodes
git clone https://github.com/dseditor/ComfyUI-ListHelper

Restart ComfyUI; look under ListHelper/Chinese. opencc-python-reimplemented installs itself on first run - if you'd rather control it: pip install opencc-python-reimplemented. ComfyUI Manager: search "ComfyUI-ListHelper".

Where people get burned

The restart is the usual stumble - run it once, watch the console, restart, then it works forever. One more thing worth knowing: this converts script, not language. Simplified/Traditional is the same language in two writing systems, so don't expect it to localize vocabulary (it won't change 「網絡」 to 「網路」 beyond what the phrase tables handle). For matching a model's expected script - which is genuinely a real issue when an LLM-encoded model's training data skews one direction - it's the right tool and the right library. Pair it with the pack's Chinese Translate node when you need to go all the way to English.