Bangtrix Simple Translate 🌐

The name undersells the useful part. This node translates text from one of 100+ languages into another - typically into English, since that's what your checkpoint's CLIP encoder was trained on - and it does it with no API key, no model download, and no setup. It's about the simplest translation node you'll find for ComfyUI, and it exists for everyone who's been copy-pasting their Japanese or Indonesian prompt into Google Translate, grabbing the result, and pasting it back into a text box. That's the workflow this node replaces.

Bangtrix Simple Translate is one of the two translate nodes shipped in the Anonymzx/BangtrixToolkit pack (the other, Bangtrix Translate Universal, does the whole translate-to-conditioning job). This one is the thin version: a string goes in, a string comes out. It never touches CLIP, never encodes anything, and it has zero model files. If you just want the raw translated text to feed into any prompt or text node, this is the one to reach for.

How it works

Don't be fooled into thinking there's a translation model hiding in this pack. The node makes an HTTP request to Google Translate's free, unofficial web endpoint (translate.googleapis.com/translate_a/single?client=gtx) using the requests library - the same endpoint a hundred free scripts have used for years. No key, no account. The practical consequences: it needs an internet connection, and it's only as fast as your network. It's not a local model.

One behavior worth knowing before it bites you: if the request fails - offline, timeout, Google throttling you - the node silently returns your original text instead of throwing an error. Your workflow keeps running, you just don't get a translation. That's the single most common "why is my prompt still in Indonesian?" moment, and the fix is usually just to check you have network access.

The inputs and outputs that matter

Only three inputs, and you'll barely touch two of them:

• text - the string to translate (multiline, so drop in a full paragraph).
• source_language - defaults to auto, which is right almost always. Set it explicitly only when auto-detection is misbehaving (rare, but it happens with short fragments).
• target_language - defaults to en, which is what you want for prompts heading into a model. The dropdown is a big list of two-letter codes, so for Chinese remember it's zh-CN (simplified) or zh-TW (traditional), not "chinese".

The single output, translated_text (STRING), wires into anything that takes text - a CLIP Text Encode node, a prompt combine node, or a Show Text node if you want to eyeball the result before it goes anywhere. It's also handy as a debugging stage: translate once, then feed the output onward.

Installing it

The pack installs like any other, via ComfyUI Manager (search BangtrixToolkit in the Custom Nodes Manager tab, install, restart) or manually:

cd ComfyUI/custom_nodes
git clone https://github.com/Anonymzx/BangtrixToolkit
pip install -r BangtrixToolkit/requirements.txt

Here's the gotcha: BangtrixToolkit is a whole toolkit, and you're getting the whole thing. Installing it also brings an auto-launching hardware-monitor overlay with its own aiohttp server, and requirements.txt pulls in aiohttp, psutil, pywin32 (on Windows), and pynvml - none of which this node needs. For Simple Translate alone, the only real dependency is requests, which almost certainly shipped with your ComfyUI anyway. If the extra baggage bothers you, that's the trade-off of a bundled pack.

Troubleshooting

• Output equals input - the silent-failure fallback kicked in. Check internet, then check whether Google is rate-limiting you (heavy use of this free endpoint can get throttled).
• It's not a local node - on an offline ComfyUI install, all you'll ever get is passthrough.
• Empty text returns an empty string, cleanly, rather than erroring.

For a one-node translation utility, that's the whole story. If you'd rather translate and enhance and get conditioning out in a single node, that's the sibling.