Deep Translator CLIP Text Encode Node

Take DeepTranslatorTextNode - the translator that lets you pick from six different backend services instead of being stuck with Google's free scraper - and fuse it with a CLIP Text Encode, and you get this node. One box: your prompt goes in in whatever language you wrote it, your chosen translation service converts it, and conditioning comes out the other side ready for your sampler.

Why you'd reach for it

The pitch is the same as GoogleTranslateCLIPTextEncodeNode - fewer nodes, one step instead of two - but with the added flexibility of deep-translator's multiple backends. If Google's free translate endpoint is unreliable for you, or you specifically want MyMemoryTranslator, or you're running a self-hosted LibreTranslate instance for privacy reasons, this node gives you that choice without giving up the one-box convenience.

How it works

Same translate-then-encode pipeline as its Google-only sibling, but the translation step goes through whichever deep-translator backend you select instead of being locked to googletrans. The translated text then runs through a standard CLIP encode against the CLIP model you provide.

The inputs and outputs that matter

• text - your prompt, multiline, in your source language.
• clip (CLIP, required) - from your checkpoint loader.
• service (default GoogleTranslator) - six choices: GoogleTranslator, MyMemoryTranslator, LingueeTranslator (word/phrase only), PonsTranslator (word/phrase only), LibreTranslator (free or api_key), QcriTranslator (free or api_key). This is the field that actually differentiates this node from the plain Google version - pick a service here based on reliability, privacy, or coverage needs.
• from_translate / to_translate - full language names, 134/133 choices, default auto → english.
• auth_data - API key or endpoint info for the services that need it (LibreTranslator, QcriTranslator beyond their free tier).
• add_proxies (bool) + proxies - route through a proxy if needed.
• Outputs: CONDITIONING (wire into your sampler) and STRING (the translated text, useful for sanity-checking with a PreviewTextNode).

How to install it

Via ComfyUI Manager: search "ComfyUI Custom Nodes AlekPet" and install. Manually:

cd ComfyUI/custom_nodes
git clone https://github.com/AlekPet/ComfyUI_Custom_Nodes_AlekPet

Restart ComfyUI - the deep-translator package installs automatically on first launch. No model downloads for this node beyond whatever CLIP model you already have loaded.

Common issues & troubleshooting

One service fails but you're not sure if it's your setup. Each backend in the service dropdown is a separate provider with its own uptime - an outage or block on one is not evidence the node is broken. Swap the dropdown to a different service before you go digging further.

LingueeTranslator or PonsTranslator returns nothing useful on a full sentence. Expected - both are dictionary-style lookups meant for single words or short phrases, not general sentence translation. Use GoogleTranslator or MyMemoryTranslator for full prompts.

LibreTranslator/QcriTranslator errors immediately. Check auth_data first - those two need a key (or, for Libre, potentially a self-hosted instance URL) unless you're relying on their free tier, which is rate-limited.

You want the translated string reused across multiple encodes (e.g. a positive and a style pass sharing the same translated base text). Same caveat as the Google-combined variant: this node translates once per call. For reuse across multiple CLIP Text Encode nodes, use the standalone DeepTranslatorTextNode once and fan its output out, rather than duplicating this combined node and re-paying for translation each time.