🌸Flower TCSC Converter

If you work with Chinese prompts - and a big chunk of this pack's audience does - you've hit the wall where a model is trained on Simplified Chinese but you write Traditional, or vice versa. A naive character swap gets you "鼠标" either way, which is wrong for Taiwan: the region says 滑鼠. Flower TCSC Converter is the pack's Traditional/Simplified bridge, and it's built on OpenCC precisely because OpenCC converts words, not just characters - so 滑鼠 ↔ 鼠标, 網路 ↔ 网络, and the other Taiwan-vs-mainland vocabulary swaps come out right.

What it actually does

Two inputs: text_input (multiline, the text to convert) and conversion_mode, a dropdown with exactly two options:

• Traditional (TW) -> Simplified
• Simplified -> Traditional (TW)

That "TW" is the whole point - the pack targets Taiwan conventions, which differ from Hong Kong and standard simplified in real, searchable ways. Behind the scenes it maps to OpenCC's tw2sp and s2twp configs (the source tries both the bare name and the .json variant to tolerate different OpenCC installs). Output is a single text_output STRING, and the node also renders the converted text in its read-only UI box so you can copy it without running the graph.

The dependency, and the one-click fix

OpenCC is the only pip dependency in the entire comfyui-flower-tools pack - it's in requirements.txt and pyproject.toml, so a normal install pulls it. But if you're on a portable Windows build where the pack's dependencies didn't install (or the import failed), the node degrades gracefully: it outputs an error telling you OpenCC is missing, and the UI shows an install button that runs the pip install for you into your Python environment and verifies it via a /flower-tools/check-opencc endpoint. Click it, confirm, restart, done. On Linux/macOS you can also just:

pip install opencc

into your ComfyUI's Python and skip the button entirely.

Where you'd use it

• Prompt localization: translate a Chinese prompt to the script your model's tokenizer handles best before encoding.
• Wildcard files: if your wildcard library is in Traditional and you're generating on a Simplified-trained model, run the selector's output through this node on its way to the CLIP encoder.
• Standalone text work: it's an output node with a preview, so it's a decent little converter you don't need a browser tab for.

Install

The whole pack installs together:

cd ComfyUI/custom_nodes
git clone https://github.com/weichenglin0215/comfyui-flower-tools.git

Then restart ComfyUI, or use ComfyUI Manager → search ComfyUI Flower Tools. Remember the OpenCC dependency above - it's the one package this pack actually needs.

Gotchas

• It's Taiwan-flavored both ways. If your target is Hong Kong Traditional (廣東話 conventions like 嘅/唔), OpenCC's tw2sp-family isn't that, and this node won't do it. Set your expectations to Taiwan.
• No batch/auto mode - one text at a time per execution. Feed it a long document and it handles it, but there's no "convert a whole folder" option.
• The conversion happens client-side in the preview too, so the node is happy to just be a converter you never wire into the graph.