CLIP Text Encoder

Functionally this is the same job as core ComfyUI's CLIPTextEncode - take a CLIP model and a string of text, output a CONDITIONING you feed into your sampler. The only reason it exists as its own node inside comfyui_LLM_party is packaging: this pack isn't purely a chat-node collection, it's built for the hybrid workflows the author actually cares about - have an LLM or VLM write or refine your image prompt, then encode it and run it straight into a normal Stable Diffusion/Flux graph without leaving the party's own node category. The README's quick-start list is full of exactly this pattern: "utilize API calls to LLM for generating SD prompts and images," "employ ollama to call minicpm for generating SD prompts and images," "utilize the local qwen-vl to generate SD prompts and images." This node is the joint where the text-generation half of the pipeline hands off to the image-generation half.

There isn't much to configure. clip is required - plug in whatever CLIP model your checkpoint loader outputs, same as you would for the stock node. text is technically optional in the schema (multiline, no default), which just means you can leave it blank and drive it entirely from an upstream wire - in practice you'll almost always connect it to the assistant_response output of an LLM node, or to a substring/string-logic node that's extracted the prompt portion out of a longer LLM reply. The single output is CONDITIONING, which goes wherever conditioning normally goes: your KSampler's positive or negative input.

Should you use this instead of the core node? Honestly, if you're not mixing LLM nodes into the same graph, just use ComfyUI's built-in CLIPTextEncode - it does the identical job and doesn't require installing this whole pack. Where this one earns its keep is purely organizational: if your workflow already has comfyui_LLM_party nodes generating the prompt text, keeping the encode step in the same category makes the graph a little easier to read at a glance, and you're not pulling in anything extra you didn't already install.

Installing it means installing the pack, since this node ships alongside everything else in comfyui_LLM_party - search "comfyui_LLM_party" in ComfyUI Manager and install, or git clone https://github.com/heshengtao/comfyui_LLM_party into custom_nodes and restart, then run pip install -r requirements.txt from inside the pack folder using ComfyUI's own Python (portable builds need python_embeded\python.exe -m pip install -r requirements.txt explicitly).

Troubleshooting-wise, there's genuinely not much unique to this node - it's a thin wrapper, so if something's wrong it's almost always upstream. A blank or garbled image usually means the LLM node feeding it produced something unexpected: check what actually landed in assistant_response before assuming the encoder is at fault, especially if your LLM sometimes wraps its answer in explanatory text ("Sure, here's a prompt: ...") rather than a clean prompt string - that's a job for the pack's Extract Substring node to clean up before it ever reaches this one.