FaceParsingProcessorLoader(FaceParsing)

This one has no settings, no inputs, and does nothing exciting on its own - and that's exactly why it trips people up when it's missing. It loads the image processor that FaceParse needs to prep pictures before they hit the segmentation model: resizing, normalizing pixel values, converting to the tensor format the model actually expects. Every vision model that comes out of Hugging Face's transformers ecosystem is paired with a processor like this, and jonathandinu/face-parsing (the model behind FaceParsingModelLoader) is no exception. Skip this node and FaceParse simply has nothing to feed the model with.

How it works

There's no configuration because there's nothing to configure - the processor is tied to the specific face-parsing model this pack ships with, so there's no choice to make. Drop the node down, and on first run it pulls the processor config from the same Hugging Face repo the model loader uses. From then on it loads locally. It's a one-node, one-job utility: instantiate the processor, hand it out as a FACE_PARSING_PROCESSOR.

The output that matters

Just one: FACE_PARSING_PROCESSOR, which plugs into FaceParse's processor input, right alongside the FACE_PARSING_MODEL from FaceParsingModelLoader and the IMAGE you're actually parsing. All three are required - FaceParse won't run without any one of them.

Installing it

Comes bundled with the rest of comfyui_face_parsing:

• ComfyUI Manager - search "comfyui_face_parsing", install, restart.
• Manual - cd ComfyUI/custom_nodes && git clone https://github.com/Ryuukeisyou/comfyui_face_parsing.git, restart ComfyUI.

If you already set up FaceParsingModelLoader and its model files downloaded fine, this one won't give you any trouble - it draws from the same repo and the same network path.

Common issues

Honestly, this node rarely breaks by itself. The one thing worth knowing: it's easy to forget to add it at all when you're wiring up a face-parsing graph for the first time, because visually it looks like it should be optional - it's a single node with no inputs sitting off to the side. It isn't optional. If FaceParse throws a missing-input error on processor, this is the node you forgot to place. Beyond that, any download hiccup on first run traces back to the same Hugging Face connectivity issue that affects FaceParsingModelLoader, and the same fix applies: grab the files manually and drop them into models/face_parsing/ if your connection to Hugging Face is unreliable.