ChatGPT Service Connector 🦉| OwlV

Why you'd reach for it

You want an actual LLM doing the translation - not a machine translator - and you want it backed by OpenAI's models. This connector is one of the three in the pack that genuinely works end to end: it packages an OpenAI API client, hands it to the LLM Translator node, and the translation happens via a live call to api.openai.com. Set it, forget it, and the quality difference over the Basic Translator node is immediate.

How it works

Under the hood it builds an OpenAIConnector pointed at https://api.openai.com/v1/chat/completions, authenticates with a Bearer token (your api_key), and sends a standard chat-completions request. When LLM Translator calls it, the response is parsed from choices[0].message.content and handed back as the translation. This is a real HTTP call with real billing - the pack just does all the request/response wrapping for you.

The inputs that matter

• api_key - required, password-masked in the UI. Your OpenAI platform key.
• model_select - the dropdown: gpt-4o, gpt-4o-mini, gpt-4-turbo, gpt-3.5-turbo, o1, o1-mini, Custom.
• custom_model_id - optional; only used when model_select is "Custom".

The output is llm, which you wire into LLM Translator's llm input, then run.

One honest note on the model list: it's a snapshot frozen into the pack. gpt-3.5-turbo has been deprecated on OpenAI's side for ages, and the list predates gpt-4.1 and gpt-5 - if the model you want isn't there, pick Custom and type its id into custom_model_id. And don't overthink the pick for translation: gpt-4o-mini is the sane default - cheap, fast, and its translation quality is genuinely good. Burning gpt-4o or an o1 reasoning model on captions is paying for horsepower you don't need. If you do try o1 and hit a refusal, remember reasoning models have stricter message rules on OpenAI's side; gpt-4o-mini is the safe fallback.

Installing it

Ships in the ComfyUI-MultiTranslator pack by OwlvChirotha ("OwlV"), installed as one unit. ComfyUI Manager → search "ComfyUI-MultiTranslator" → Install → restart. (The README's "search ComfyUI-Translator" instruction is stale - that's the old repo name.) Or by hand:

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

Restart, and the node appears under Text Processing/LLM. Dependencies are just googletrans and requests - no model downloads, because the model lives at OpenAI.

Troubleshooting

• A 401 (bad or expired key) surfaces as an error string in the translated output rather than a crash - check the node output.
• The key is masked in the UI, but it still lives inside the workflow JSON. Clear it before sharing workflows; the README makes the same point.
• Every translation run is a full API call of your entire text. If you're doing lots of runs, keep the input texts short or you'll notice the bill.