Ino Hugging Face Download Model

Every serious ComfyUI user eventually hits the wall of "I need a model that isn't downloaded yet," and the usual fix is a terminal, a browser tab, and a whole lot of hoping the file lands in the right folder. Ino Hugging Face Download Model moves that into the graph: you give it a repo and a filename, and it downloads the file into the correct ComfyUI model directory, reporting the exact path it landed at. Re-run the workflow and the model's just... there.

It's part of the Ino Nodes pack from nobandegani, and it's the workhorse of the pack's download layer. It's built on huggingface_hub (with the pack's hf_xet dependency for the newer transfer transport), so it downloads the same way the official tooling does.

What you set

The required inputs are the download coordinates:

• repo_id - the HuggingFace repo, e.g. black-forest-labs/FLUX.1-dev.
• filename - the file inside that repo to fetch.
• model_type - dropdown; decides which ComfyUI model folder the file goes into (text_encoders, vae, diffusion_models, etc.).
• model_subfolder - subfolder under that model directory (defaults to flux1dev).
• subfolder - optional subfolder within the repo where the file lives.
• model_config - an alternative way to supply all of the above: a config JSON string. If you've built a config with the pack's config nodes, you can feed it here and skip the individual fields.

Optional extras you'll rarely need: token (for gated repos), repo_type (for non-model repos like datasets), revision (a specific commit or branch), and ignore_repo_dir.

Outputs: success / msg for status, plus model_type, abs_path, and rel_path telling you exactly where the file landed.

How it works

Fill the fields (or feed a config), and the node resolves the destination from ComfyUI's own model folders, then calls huggingface_hub to stream the file down. Big models stream in chunks, so you can watch the download happen rather than freezing the UI. The path outputs are the useful part - wire abs_path into a loader, a hash-check node, or a file counter to build download-verify-load pipelines without ever leaving the canvas.

The honest caveats: it downloads a single file - for whole repos there's a separate Ino node (Ino Hugging Face Download Repo). Gated models need your token. And while it fetches files, it doesn't load them; pair it with a loader node when you want the model usable right away.

Installing it

It ships with the whole Ino Nodes pack. ComfyUI Manager: search "ComfyUI Ino Nodes", install, restart. Or manually:

cd ComfyUI/custom_nodes
git clone https://github.com/nobandegani/ComfyUI-InoNodes.git
cd comfyui_ino_nodes
pip install -r requirements.txt

Restart after. Pack-wide caveats apply: it's built on ComfyUI's V3 schema (README wants v0.18.1+), and requirements.txt installs inopyutils, plus huggingface_hub and hf_xet, which are exactly what this node needs.

Common issues

A 404-style failure almost always means the repo_id/filename/subfolder combo is wrong - double-check the exact filename in the repo's file browser. If the file is huge and seems stuck, it's a big download, not a hang; large diffusion models legitimately take minutes. And remember the destination folder is decided by model_type: a file named as a VAE won't land in the VAE folder unless model_type says so. Search "Ino Hugging Face Download Model" in the node menu (grouped under InoModelHelper).