FC StyleLoraLoad

FC StyleLoraLoad is the "load a whole style, fully configured" node in ComfyUI-FaceChain - and this page is for the variant of that node registered with the dash-slug spelling. There are two pages for the same underlying node on comfy.icu (FC_StyleLoraLoad and this one) because the pack's class name contains a space (FC StyleLoraLoad), which different crawls and workflow exports render with different slug separators. Same node, same behavior, one dropdown.

So if you've read the FC_StyleLoraLoad page, you already know everything. The short recap for people landing here:

What it does

One input - style_name, an enum of ~75 preset styles. It loads the matching base checkpoint (either leosamsMoonfilm_filmGrain20 or MajicmixRealistic_v6), merges the style's LoRA in at the intended strength, and outputs:

• MODEL, CLIP, VAE - wire straight into your sampler, exactly like a checkpoint loader's outputs
• style_prompt (STRING) - the style's canonical prompt, for a CLIP Text Encode node

The whole point is that you don't have to hunt down "the right" checkpoint and LoRA for a FaceChain style - the pack ships the style definitions in its styles/ directory, and this node resolves model, LoRA, strengths, and prompt for you.

The catch that matters

Everything downloads from ModelScope (Alibaba's hub), not Hugging Face. First use of a base model runs snapshot_download for the checkpoint and the style LoRA, and if the checkpoint isn't already in your ComfyUI checkpoints folder, it lands there as leosamsMoonfilm_filmGrain20.safetensors or majicmixRealistic_v6.safetensors. That's a multi-gigabyte first-run download, and if your network can't reach modelscope.cn the node stalls silently. Also expect the checkpoint-load-plus-LoRA-merge to be slower than a plain loader - that's the cost of the convenience.

If ComfyUI says the node is missing

Because the class name has a space, some saved workflows reference it with different slug conventions. If a loaded workflow complains, re-select the style in the node or re-save the workflow - the underlying node is the same FC StyleLoraLoad in both slugs, so there's no separate install.

Install

Standard pack install: ComfyUI Manager (search "ComfyUI-FaceChain") or:

cd ComfyUI/custom_nodes
git clone https://github.com/THtianhao/ComfyUI-FaceChain

Restart, and give the first launch time - the pack auto-installs its heavyweight dependency stack at startup (onnxruntime-gpu, mmdet==2.26.0, mmcv, modelscope, insightface, transformers…) before the style dropdown even renders. Then budget for the ModelScope model downloads on your first style selection.

The one-line verdict: it's the pack's most opinionated convenience node - genuinely nice when you want a FaceChain style without assembling the parts, with the ModelScope-first download behavior as the price of admission.