Text Translation V2

The plain Text Translation node does one thing well: anything you feed it comes out in English, via Bing, no questions asked. Text Translation V2 is the version that asks questions. You get to choose the engine - Bing, Google, or Alibaba - and you get to choose the source and target languages instead of being stuck with auto-detect-to-English. Same zero-key, zero-account deal, same translators library underneath, just with the dials exposed.

When it's the one you want

Most of the time, honestly, the plain node is enough: auto-detect → English is the right default for feeding a prompt to a CLIP/LLM text encoder. But you reach for V2 when:

• Bing is rate-limiting or flaky from your region and Google behaves better. They're not interchangeable in practice.
• You want English → your own language, not the other way around. Handy when you download a workflow whose prompts are in a language you don't read, or when you're translating a character prompt to check it says what you think it says.
• You're translating into a language that isn't English and the plain node can't do that at all.

The inputs that matter

• trans_switch - boolean, on by default. On translates, off passes text through untouched. Same bypass trick as the plain node.
• trans_text - the multiline text you want translated.
• translator - Bing, Google, or Alibaba. Bing is the default, matching the plain node's behavior.
• source_language - auto plus eleven languages. Keep it on auto unless you know better.
• target_language - eleven languages, with English(en) first.

The output is a single string, wired into a CLIP Text Encode or wherever the text goes next.

Two mechanics worth knowing. First, the languages are stored as display strings like 中文(zh); the node parses out the code before calling the library, so what you see is what you pick. Second, the node ships with a front-end "Show / Hide button" - by default it hides the extra dropdowns behind a click, so a fresh node looks nearly as clean as the plain one until you need the controls.

Installing and the known error

Same install as the rest of the pack - ComfyUI Manager (search ComfyUI-Text-Translation), or:

cd ComfyUI/custom_nodes
git clone https://github.com/pictorialink/ComfyUI-Text-Translation
cd ComfyUI-Text-Translation
pip install -r requirements.txt

Restart, and remember the pre-May-2024 versions are incompatible with the current pack - delete the old folder before updating.

This node is literally the one the README shows in its error section:

Prompt outputs failed validation
Text_Translation_V2:
- Value not in list: translator: 'alibaba' not in ['Bing', 'Google', 'Alibaba']

After a pack update, an existing node can hold a stale widget value (say, a translator that no longer exists, or a lowercase one). It's not your fault and not your config. Delete the node and re-add it. If you're switching engines a lot, expect the occasional rate-limit hiccup from whichever endpoint you picked - that's the price of free.