Google Translate2 (CLIP Text Encode)

This is the CLIP-encode version of the "2" family: type your prompt in whatever language, get auto-detected-and-translated-to-English conditioning out the other end. The "2" distinguishes it from the AlekPet-style GoogleTranslateCLIPTextEncode node in the same pack - same free googletrans engine, but with no language dropdowns at all. It's hard-coded to detect the source language and translate to English, which is exactly the 90% case, and it encodes the result for you so you never have to touch the translated text unless you want to.

Free, no API key, and it does the whole prompt→English→conditioning pipeline in one node. That's the pitch, and it's a good one for non-native English speakers.

How it works

Same skeleton as the other "2" node: googletrans async translation (src="auto", dest="en"), then a CLIP encode. The notable difference under the hood is how it encodes: this node uses clip.encode_from_tokens_scheduled(tokens) rather than the standard encode_from_tokens. Scheduled encoding is the variant typically associated with AnimateDiff-style conditional scheduling, but it works fine as a general encode here - you don't need to do anything differently; it just means the conditioning it emits is built the same way the core animated workflows build theirs. Wire the CONDITIONING output into your sampler's positive or negative port.

Inputs: text_to_translate (multiline) and clip (from your checkpoint loader). Outputs: CONDITIONING and translated_text (STRING) - the string is a free bonus so you can see what it actually sent to the encoder.

Honest caveats

Same free-translation family tradeoffs: it's a scraper, so it can be rate-limited and occasionally returns an error string ("Translation error: ...") instead of a translation; it needs network on every run; and auto-to-English means there's no way to translate to any other language - if you need French out, this is the wrong node, and the non-"2" CLIP node in this pack is the one with the dropdowns. Also - mechanical translator: weighted prompt syntax like (word:1.2) or dense tag lists can get reworded into something less precise, so for prompts that depend on weights, translate the prose and add the weights back by hand. And a practical one: if you already have AlekPet's ComfyUI_Custom_Nodes_AlekPet installed, you probably have the non-"2" equivalent under a different menu and don't need this duplicate.

Install

ComfyUI Manager: search ComfyUI_Gayrat. By hand:

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

Restart ComfyUI; it's under Gayrat/translate. No models, no key - just the pack's googletrans dependency and your existing checkpoint's CLIP.