Diffusers Model Loader

This is the quieter sibling of Diffusers Hub Model Down-Loader - same underlying idea, no download step. Where that node fetches a model from HuggingFace by repo ID, Diffusers Model Loader points at a diffusers-format model you've already got sitting on disk (cloned manually, downloaded earlier, or exported from a training run) and just loads it. If you're re-running the same model repeatedly, or you're offline, or you'd rather manage the download yourself, this is the one you want instead.

Why "diffusers format" is the detail that matters

HuggingFace's diffusers library expects models laid out as a pipeline folder - separate subfolders for the UNet, text encoder(s), VAE, and scheduler config, plus a model_index.json tying it together - not the single .safetensors file most ComfyUI checkpoints ship as. WAS built its own path for loading that layout directly, which is why this exists as a distinct node rather than something the standard checkpoint loader handles. If a model you want is only distributed in that multi-file diffusers layout, this is how you get it into a WAS graph without hand-converting it first.

How it works

You give it a local path to a diffusers-format model directory, and it loads that pipeline into memory for downstream WAS nodes to use. It doesn't fetch anything from the network - that's the entire difference from its Hub-downloading sibling. Because the object it produces isn't ComfyUI's native MODEL/CLIP/VAE triple, it's meant to feed WAS's own diffusers-aware nodes rather than a standard KSampler chain built for checkpoint-loaded models.

When you'd reach for this over the Hub loader

Anytime the model's already on your machine and you don't want to re-download it every session, or you're working somewhere without reliable network access to HuggingFace, or you manage your model downloads separately (a shared model drive, a script, whatever) and just need ComfyUI to pick up what's already there.

How to install it

ComfyUI Manager: search "WAS Node Suite," install, restart. Manual:

cd ComfyUI/custom_nodes
git clone https://github.com/WASasquatch/was-node-suite-comfyui

then install requirements from inside that folder - path/to/ComfyUI/python_embeded/python.exe -s -m pip install -r requirements.txt on portable, pip install -r requirements.txt on system Python. Restart ComfyUI; find it under WAS Suite.

Common issues & troubleshooting

The most likely failure is a path or layout problem, not a bug in the node: if the folder you point it at isn't a genuine diffusers pipeline directory (missing model_index.json, or it's actually just a loose .safetensors file), loading will fail. Make sure whatever you downloaded is the full diffusers export, not a single-file conversion - those need the standard checkpoint loader instead, not this node.

Beyond that, this inherits the pack's general risk profile: WAS Node Suite has had no active development since it was marked retired in December 2023, and a loader built against a particular version of the diffusers library is exactly the kind of thing that can quietly break as that library moves on, with nobody upstream chasing the fix. If the whole suite goes "Import Failed" after a ComfyUI update, that's usually a dependency version clash rather than anything about this specific node - re-running the requirements install with your embedded Python interpreter (not a system one) is the fix that's actually worked for people stuck on it.