unCLIP Checkpoint Loader

unCLIP checkpoints are a specific, older branch of Stable Diffusion (Stable Diffusion 2.1-unCLIP is the well-known example) built to accept a CLIP image embedding as conditioning alongside - or instead of - text. That's what makes them different from an ordinary checkpoint: feed one a reference image, and it can generate variations of that image's content rather than starting purely from a text prompt. ComfyUI ships a core loader for this family; WAS Node Suite's version does the equivalent load and then, following the same pattern as its plain Checkpoint Loader, adds a text output naming the checkpoint file you loaded.

What it's for

If a workflow specifically calls for an unCLIP checkpoint, this is the node that gets it into your graph - loading the model pieces an unCLIP-conditioned pipeline needs and reporting back which checkpoint file supplied them, useful for the same reason it's useful on WAS's regular loader: you can wire that filename into a save pattern or caption without hardcoding it separately and having it drift out of sync when you swap models.

Worth saying plainly, as a heads-up rather than a knock on the node: unCLIP checkpoints are a niche, aging corner of the ecosystem at this point. Most workflows that want image-conditioned generation today reach for IP-Adapter instead - a separate, actively maintained approach to reference-image conditioning that works across a much wider range of modern checkpoints, not just the specific unCLIP-trained ones. Reach for this loader when a workflow specifically names an unCLIP checkpoint; if you're starting a new image-conditioning setup from scratch, IP-Adapter is worth checking first.

How it works

Same shape as any checkpoint loader: point it at a file, and it loads what an unCLIP pipeline needs to run - including the CLIP vision component that plain SD checkpoints don't carry, which is what lets image embeddings feed into conditioning in the first place - plus the filename string. Point it at a checkpoint that isn't actually an unCLIP model and expect it to fail or behave incorrectly, since it's expecting that specific weight layout.

Installing it

Bundled with the full WAS Node Suite pack:

• ComfyUI Manager - search WAS Node Suite, install, restart.
• Manual - cd ComfyUI/custom_nodes && git clone https://github.com/WASasquatch/was-node-suite-comfyui/, then pip install -r requirements.txt from inside that folder, then restart ComfyUI.

This node loads whatever checkpoint file you point it at from your standard checkpoints folder - it doesn't fetch anything on its own.

Common issues

The most likely source of confusion is pointing this loader at a regular (non-unCLIP) checkpoint and getting errors or nonsense output - the fix is making sure the file you're loading was actually trained as an unCLIP model, not just any SD checkpoint.

Beyond that, the same pack-wide caveat applies here as everywhere else in WAS Node Suite: it's been unmaintained since December 2023, and "Import Failed" errors after a ComfyUI update are a known, recurring complaint tied to the pack's older pinned dependencies. Reinstalling requirements.txt inside your ComfyUI venv is the usual fix if the whole suite stops loading.