ClipClamp

ClipClamp is a pass-through node from the Allor plugin. It takes a CLIP object in, hands the exact same CLIP object back, and does absolutely nothing in between. If that sounds useless, you're half right - but "nothing" is exactly the feature, and this is one of the most underrated graph-hygiene tricks in ComfyUI.

CLIP is the text-encoder object that a checkpoint loader or a dedicated CLIP loader hands to your CLIPTextEncode node. Normally you wire it straight through and never think about it again. So when would a no-op that sits in that wire help?

• Splitting one CLIP to several destinations. A loader output can feed a dozen encoders, but the wires get visually chaotic fast. Run the CLIP through one ClipClamp and fan out from its output - or chain clamps to label the "branches" of your graph.
• Forcing a stable reference point. A pass-through node creates a real dependency edge in the graph, so anything hanging off the clamp's output is guaranteed to be scheduled after the clamp. In big workflows where execution order quietly bites you, that little forced ordering can save a debugging session.
• Keeping shared workflows tidy. When you save a workflow as JSON to share, an explicit named clamp reads a lot better than a loader output wired invisibly into eleven places.

It's a one-in-one-out node with no parameters. Input is clip (CLIP), output is CLIP. That's the entire schema - nothing to tune, nothing to break.

Where it comes from

ClipClamp is part of Allor ("Allor Plugin"), Nourepide's 90+ node image-processing pack. Allor ships a whole family of these clamps - ClipVisionClamp, ConditioningClamp, ControlNetClamp, GligenClamp, plus clamps for MODEL, MASK, LATENT and IMAGE - all with the same pass-through design. The pack's pitch is doing operations in tensor space to avoid needless conversions, and this family is the boring-but-useful end of that: cheap, safe, no data transformation at all.

Installing it

Install the whole pack from ComfyUI Manager (search "Allor"), or clone it manually:

cd ComfyUI/custom_nodes
git clone https://github.com/Nourepide/ComfyUI-Allor
cd ComfyUI-Allor
pip install -r requirements.txt

Then restart ComfyUI. One heads-up: the pack's requirements.txt pulls in rembg and onnx because of its segmentation nodes, even though ClipClamp itself needs nothing - it's pure Python pass-through. Also, Allor's repo was rebased to strip images from git history (it's ~344x smaller now), which can make auto-updates misbehave; if Manager reports update errors, the pack's docs at nourepide.github.io/ComfyUI-Allor-Doc have a dedicated troubleshooting page for exactly that.

Real talk

You don't need ClipClamp. ComfyUI lets you wire one output to many inputs directly, and a no-op node is, on paper, wasted keystrokes. But once you've inherited a 200-node workflow with wires crossing the entire canvas, you'll see the appeal. Use it as a glorified label, a forced ordering point, or a visual separator. Just don't expect it to fix anything - its whole job is to not be the thing that's broken.