ChineseToCharacter

A lookup table disguised as a node. Type an anime character's Chinese name and ChineseToCharacter returns the corresponding English tag set - gender, character name, and series - formatted the way tag-based models like to see it. It's built specifically around Animagine XL, whose prompt style and character vocabulary this pack's author was clearly working with, and it saves you from guessing the English spelling of a character whose name you only know in Chinese.

How it works

The node reads a CSV file bundled with the pack - animagine-xl-3.1-characterfull-zh.txt - that maps columns for gender, English character name, English series, Chinese character name, and Chinese series. Your chinese_name input is matched against the Chinese character column (substring match), and on a hit the node returns a single string like:

female, Hatsune Miku, Vocaloid

That's gender, character_en, anime_en joined with commas - exactly the tag prefix you'd paste into an Animagine-style prompt. No match? You get the literal string Character not found.

Inputs and outputs

One input: chinese_name (STRING, default "填写角色中文名", i.e. "fill in the character's Chinese name"). One output: text (STRING) - the tag string, or Character not found.

The big gotcha

This node only works if the data file exists - and it doesn't in the current repo. The April 2025 rewrite that slimmed the pack down to just the GPT4V captioner also deleted animagine-xl-3.1-characterfull-zh.txt along with the whole "Sanmi Simple Nodes" code file. So on a fresh install the node won't appear, and even if you resurrect the code from git history you still need that CSV next to it, or the pandas read will fail on import. If you're set on using it, grab both the old Load_Image_Batch.py and the data file from the repo's pre-rewrite commits (e.g. b4487aab).

What it's not

It's a static lookup, not a translation service and not a model - no API, no ML, nothing to download, and nothing fuzzy. If a character isn't in that one CSV, you get Character not found and no fallback. For mainstream anime tagging you're honestly better off with WD14 tagger or just knowing your Danbooru tags, but for the specific case of "I know this character's Chinese name and I need Animagine's English tags," this does exactly one thing and does it instantly.

Install

cd ComfyUI/custom_nodes
git clone https://github.com/438443467/ComfyUI-GPT4V-Image-Captioner

Restart ComfyUI. No models. Remember: you'll need the deleted CSV restored for it to function - this is one of the most fragile nodes in the pack for exactly that reason.