DeepSeek Service Connector 🦉| OwlV

Why you'd reach for it

DeepSeek is the README's own recommendation for mainland-China users: no proxy gymnastics, an OpenAI-compatible API, and famously low prices. If the Basic Translator's Google path is blocked on your network, this is the LLM-grade alternative that works from a China network without extra hops. And even outside that constraint, DeepSeek is one of the best price-per-quality deals around for plain translation work.

How it works

The node builds a DeepSeekConnector pointed at https://api.deepseek.com/chat/completions, authenticates with your api_key as a Bearer token, and sends a standard chat-completions request. When the LLM Translator node calls it, the reply is read from choices[0].message.content. It's a real, live API call - the same OpenAI-style shape DeepSeek's API exposes, so this is one of the genuinely working connectors in the pack.

The inputs that matter

• api_key - required, password-masked. Your DeepSeek platform key.
• model_select - deepseek-chat, deepseek-coder, deepseek-v3, deepseek-r1, Custom.
• custom_model_id - optional; used when model_select is "Custom".

Output is llm, which plugs into LLM Translator's llm input.

Model picks, briefly: deepseek-chat (the default) is the right choice for translation. deepseek-coder is a code model - wrong tool for prose. deepseek-r1 is a reasoning model; it can overthink a simple sentence, and you're paying for tokens it spends thinking. For translation, stay on chat or v3.

Installing it

Ships in the ComfyUI-MultiTranslator pack by OwlvChirotha ("OwlV"). Install the whole pack: ComfyUI Manager → search "ComfyUI-MultiTranslator" → Install → restart. (Ignore the README's "ComfyUI-Translator" search string - stale.) 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 look under Text Processing/LLM. Dependencies are just googletrans and requests; the model itself lives on DeepSeek's servers.

Troubleshooting

• Errors surface as the translated output text, not a crash - a 401 means the key is wrong or expired.
• The key is masked in the UI but still stored in the workflow JSON; clear it before sharing.
• DeepSeek is rate-limited per API tier; if you're hammering long caption batches, you'll hit limits faster than with the more expensive providers. Batch deliberately.