Basic Translator 🦉| OwlV

Why you'd reach for it

You grabbed a workflow by a Chinese creator and every prompt node is in 中文. Or your own prompting strategy runs in your native language, but CLIP wants English. Or you're captioning a dataset and want a quick machine pass before a human touches it. Basic Translator is the "no LLM, no thinking, just translate it" node: give it a string, pick a language, and it hands back a translated string you can wire straight into a CLIP Text Encode, a text save node, or anything that eats a STRING.

It's the one node in the OwlvChirotha/ComfyUI-MultiTranslator pack that needs no API key to get started - and, as you'll see, one of the few that's actually finished.

How it works

Four services live under one hood:

• Google Translate - free, via the googletrans library. No key, no config. It's an undocumented web endpoint, so think "best effort," and it wants an international network. The README is blunt about it: "Requires proxy in mainland China."
• Baidu Translate - the official API, MD5-signed request. You register at fanyi-api.baidu.com for an App ID and secret key, and get roughly 2 million free characters a month. This is the README's recommendation for mainland-China users.
• Youdao Translate - official API with SHA-256 signing, keys from ai.youdao.com, daily free quota.
• Tencent Translate - the honest warning: it isn't implemented. The code's Tencent path just returns a message telling you to use Baidu or Youdao instead. Don't go hunting for a Tencent Cloud SDK setting that doesn't exist in the UI.

Each service speaks its own language codes, and the node maps your friendly dropdown choice ("Chinese", "Japanese", …) to the right code per service. The source language is auto-detected, so you never tell it what you're translating from.

The inputs that matter

• text - multi-line; type it or wire it from an upstream node.
• target_language - ten options: English, Chinese, Japanese, Korean, French, German, Spanish, Italian, Russian, Portuguese.
• translator_service - pick Google, Baidu, or Youdao (and technically Tencent).

The optional inputs are only for the keyed services: baidu_app_id + baidu_secret_key, youdao_app_id + youdao_secret_key, and the moot tencent_secret_id + tencent_secret_key.

The output is translated_text, a plain STRING - your hook into the rest of the graph.

Installing it

This node ships in the ComfyUI-MultiTranslator pack by OwlvChirotha ("OwlV"), and the whole pack installs at once. ComfyUI Manager → search "ComfyUI-MultiTranslator" → Install → restart. (The README tells you to search "ComfyUI-Translator" - that's the project's old name and the string is stale. Use "ComfyUI-MultiTranslator".)

Or by hand:

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

Then restart ComfyUI. The dependency list is refreshingly small - just googletrans and requests, no model downloads, no heavy extras. You'll find the node under Text Processing in the node menu.

Troubleshooting

• Pick Baidu or Youdao without filling in the keys and you don't get a crash - you get the translated_text output filled with an English error string. Watch the output text, not the console.
• Google returns garbage or errors: that's the free endpoint being flaky or blocked on your network. Switch to Baidu/Youdao with keys, or step up to the pack's LLM translator.
• Empty input returns an empty string, so you can chain the node without a guard for the empty case.

That's the whole thing. A small, honest node: three real machine translators, one placeholder, and zero friction on the Google path.