ER Model Downloader

The universal ComfyUI onboarding ritual: you download someone's gorgeous workflow, drag it in, hit Run, and get an error naming a model you've never heard of. Then you open HuggingFace, search by a filename that may or may not match, download it to the wrong folder, and discover your error was for a different missing model all along. ER Model Downloader is that ritual, automated: it scans the current workflow, lists every model file it uses, tells you which subfolder each belongs in, flags what's missing, and downloads it to the right place with a progress bar.

It's a utility node from the ER VFX Nodepack by Héctor Gallego (ER Academy), and it's the pack's answer to workflow portability. It has no inputs and no outputs - it's pure front-end + server plumbing. You drop it anywhere in the graph, press 🔍 Scan workflow, and read the list.

What the scan tells you

For every model the workflow references - checkpoints, LoRAs, VAEs, text encoders, diffusion models, ControlNets, all of it - it reports:

• which models/ subfolder the file belongs to (checked against your real model folders),
• whether it's already installed or missing.

For each missing one you then have three ways in, in order of convenience:

1. Embedded URL - if the workflow itself carries a download URL (modern template workflows do), it's pre-filled for you.
2. 🔎 HF - searches HuggingFace by the exact file name and fills in the best match's URL.
3. Manual URL - paste any https:// link.

Then ⬇ downloads into the selected folder with a live progress bar, or ⬇ Download all missing handles the whole list. Downloads are resume-safe: files write as .part and only get renamed when complete, so an interrupted download restarts cleanly instead of corrupting. You can change each file's target folder via dropdown before pulling it down.

The gotchas the docs flag

• Gated HuggingFace models (401/403) need a token. Set the HF_TOKEN environment variable before launching ComfyUI, or download that one manually. This bites people with FLUX and other gated repos constantly.
• Refresh your loader combos after downloading - press R in ComfyUI or the new files won't appear in the dropdowns (that's the standard refresh shortcut for node combo lists).
• Only https:// URLs and real ComfyUI model folders are accepted by the server endpoint, deliberately - no file:// shortcuts, no random paths.

What it won't do

It doesn't download non-model requirements (Python packages, other custom nodes), and it can't fix a workflow whose model files genuinely don't exist publicly under the referenced name - the HF search does its best by exact filename, but an obscure renamed checkpoint may still need a manual hunt. It's a scanner and a downloader, not a dependency resolver.

Installing

ER Model Downloader ships as part of the ER VFX Nodepack. ComfyUI Manager → search "ER VFX Nodepack" → install, or:

cd ComfyUI/custom_nodes
git clone https://github.com/esperandoelrender/ComfyUI-ER-VFX-Nodepack.git

Restart and find it under utils. No extra Python dependencies (it uses ComfyUI's own folder registry and aiohttp server plumbing). If you previously installed the standalone er_model_downloader pack this grew from, remove it first - this pack supersedes it.

My honest take: for the specific pain of "workflow I didn't build is missing models," this is the smoothest in-pack utility there is - and it's a good habit to drop it into any downloaded workflow and scan before your first run, so the error list comes from the node instead of the console.