Hugging Face Download Folder
Grab a whole Hugging Face folder (or repo) in one node
- model_path
Sometimes the thing you want isn't one file - it's a folder. A quantized GGUF set that ships five files per model, a repo that bundles its CLIP and VAE next to the checkpoint, a LoRA collection. Hugging Face Download Folder exists for exactly that: same pack as the sibling Hugging Face Download Model node, same two inputs, but instead of grabbing a single file it recursively pulls a whole folder - or an entire repo - into a ComfyUI model folder.
How it works
Paste a Hugging Face link and the node resolves it into repo + subfolder. On run it downloads everything under that remote path into your chosen target_folder. Point it at a folder URL (a tree link) and it fetches that folder; give it just a repo and it pulls the whole thing, naming the local folder after the repo's last segment - openai/clip-vit-large-patch14 lands in a clip-vit-large-patch14 folder. It shares all the plumbing with its sibling: the same link parser, the model-cache clear, and the dropdown refresh after a successful download.
One real difference from the file node: folder mode is Hugging Face-only. The non-HF direct-URL fallback that the model node supports doesn't apply here - if your file isn't on the Hub, this node won't help.
The model_path output is a string with the folder's relative path (note the trailing slash when it targets a standard model folder) - handy for wiring into anything that wants a folder, or just to confirm where things landed.
The inputs
Identical to the model node, and that's the whole list:
- link - a repo URL, or a
tree/subfolder URL to scope the download. - target_folder - dropdown of your model subfolders (
checkpoints,loras,diffusion_models, …), orcustomto type a path.
Install and gotchas
Install exactly like its sibling - ComfyUI Manager search for "Hugging Face Download & Backup", or manually:
cd ComfyUI/custom_nodes
git clone https://github.com/jnxmx/ComfyUI_HuggingFace_Downloader
cd ComfyUI_HuggingFace_Downloader
pip install -r requirements.txt
then restart. Same caveats apply: comfy-cli rides along in the requirements, downloads are synchronous so a big repo will freeze the queue while it runs, and gated repos need a token (downloader.hf_token in settings or HF_TOKEN).
And the one warning that matters most for this node: before you download a whole repo, check whether you actually need everything in it. Some HF repos bundle a hundred files you'll never load, and this node will cheerfully grab them all. Point it at the specific subfolder you want and save yourself the download time.
Inputs (2)
| Name | Type | Default | Description |
|---|---|---|---|
| target_folder | COMBO | 29 options: [object Object], [object Object], [object Object], [object Object], [object Object], [object Object], +23 | |
| link | STRING | — |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| model_path | * | — |