Load LLM Model 🦉| OwlV

Why you'd reach for it

You want translation that never leaves your machine: no API key, no per-call billing, your prompts staying local. That's the promise of "Load LLM Model", and it's the right instinct - local LLMs are the private, free-once-you-own-the-hardware path, and the ecosystem has gotten very good at running them on consumer GPUs (GGUF quantized models are the usual route). So a node that loads a local LLM for translation should be a no-brainer.

Here's the thing the README won't tell you: in the shipped code, this node is a stub.

How it actually works

The load method doesn't load anything. It checks that a directory exists, then returns an LLM client whose translate() produces a placeholder string like [Local /path/to/models/LLM] -> 中文: your text. There's no inference, no GGUF loading, no llama.cpp under the hood - the source literally carries a # TODO comment for implementing local model loading and inference. Wire it into the LLM Translator node and you get that placeholder string back, not a translation.

There's also a path quirk worth knowing. The dropdown option reads "ComfyUI models/LLM", but the code computes the folder as <comfyui>/custom_nodes/models/LLM - a sibling of the pack's own folder, not ComfyUI's actual models/LLM directory where checkpoints and GGUF files normally live. So even the "correct" default points somewhere that usually doesn't exist on a stock install.

It's not useless - it's shape-compatible, so you can build and test a workflow graph with it while you wait for a real backend. But "loads a local LLM" it does not.

The inputs

• source - "ComfyUI models/LLM" or "自定义目录" (custom directory). Half the dropdown is in Chinese; that's the pack's bilingual UI.
• custom_path - the absolute path, used when source is the custom directory.

Output: llm, which feeds LLM Translator's llm input.

Installing it

Ships in the ComfyUI-MultiTranslator pack by OwlvChirotha ("OwlV"). ComfyUI Manager → search "ComfyUI-MultiTranslator" → Install → restart. (The README's "ComfyUI-Translator" search string is 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, then look under Text Processing/LLM. Dependencies are just googletrans and requests - notably, nothing that could actually run an LLM. That's the tell.

If you actually want local LLM translation

Don't wait on this one. The ComfyUI ecosystem has proper local-LLM nodes that load GGUF or Q4/Q8 quantized models and actually run inference - that's the mature path for local translation. Use one of those for real work, and keep an eye on whether OwlvChirotha finishes this loader behind its existing llm interface. When they do, this node will slot into workflows that were built against it. Until then, it's a placeholder with a confident name.