转化为utf-8编码 / Convert to UTF-8

Somewhere between copy-pasting text from random sources and shoving it through a pipeline, encoding can go sideways - lone surrogates, orphaned bytes, characters that are technically there but unreadable. UTF8 Converter is the cleanup node for that mess: it strips out everything that isn't valid UTF-8 and hands you back a clean string, plus a log of exactly what it removed.

The realistic use case is pre-processing text before it hits something strict - an API call that rejects malformed JSON, a save step that chokes on weird bytes, a subtitle or TTS stage that garbles mojibake. If you've ever had an LLM call fail with a cryptic encoding error and spent an hour hunting a phantom quote character, this is the node that would have found it in one pass. It's the "sanitize before you send" step that nobody builds until they need it once, and then they need it constantly.

How it works

Feed it input_text (multi-line) and it filters out invalid sequences, returning filtered_text plus a log string that details what was removed and a summary - useful both for debugging and for verifying the cleanup did what you think it did.

Two inputs control behavior:

• remove_emoji - default true, and here's the trap: it force-deletes all 4-byte characters, which includes the vast majority of emoji. If your text contains emoji you want to keep, flip this off.
• input_file - an optional file path. If you fill it in, the node reads and processes that file instead of the input_text box. Handy for cleaning up a text file on disk without pasting its entire contents into the UI. The node even detects binary files and handles them gracefully rather than producing garbage.

Outputs: filtered_text (STRING) and log (STRING).

Installing it

Part of the RUI-Nodes pack ("Rui-Node🐶"). Install once via ComfyUI Manager (search "RUI-Nodes") or:

cd ComfyUI/custom_nodes
git clone https://github.com/rui40000/RUI-Nodes
cd RUI-Nodes
pip install -r requirements.txt

Restart after. Pure Python - no extra dependencies.

Where people get burned

The emoji default is the one that catches people: remove_emoji being on by default means your "clean" output silently loses all emoji, which is great for API-bound text and awful if your text's meaning lived in those characters. Check the log - it tells you exactly what got dropped. Second, understand what it does not do: it doesn't re-encode mojibake back into proper characters, it deletes what's broken. Text that's been double-encoded will still look wrong after cleaning; the node just guarantees the result is valid UTF-8, not that it's the text you originally wanted. For most pipeline-cleaning jobs that's the right contract.