ControlNetClamp

ControlNetClamp is the pass-through node for CONTROL_NET in the Allor plugin. It takes a CONTROL_NET object in and returns it untouched - no merging, no strength adjustment, no re-encoding. If you're new to this family, that sounds useless; if you've fought a 200-node workflow, you know exactly why it exists. It's graph hygiene, pure and simple.

CONTROL_NET is the object you get from a ControlNetLoader - the baked weights that a ControlNet apply node (like ControlNetApplyAdvanced) uses to steer generation from an image, depth map, or pose. Normally the wire is simple: loader → apply. But in a serious workflow you often end up with the same ControlNet feeding multiple stages - a Canny pass at one strength plus a second application with different conditioning, or one ControlNet shared across a tiled/regional setup. That's where a clamp is worth dropping in:

• A clean fan-out point. Run the loaded model through ControlNetClamp and hang every apply node off its output. One visible split instead of a loader output stretched across half the canvas.
• Forced scheduling. The clamp is a real node, so anything downstream is guaranteed to execute after it. If a ControlNet apply occasionally runs before its loader finishes in your graph, this gives you a deterministic ordering point.
• Readable saved workflows. ControlNet is fiddly and people load each other's JSONs to steal setups - which is culture, per the knowledge base. An explicit "controlnet branch" node makes the structure legible at a glance.

The node itself has zero knobs: control_net_clamp in, CONTROL_NET out. You cannot misconfigure it, which is precisely the point.

Installing it

It ships with Allor:

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

Restart ComfyUI, or install "Allor" via ComfyUI Manager. Two pack-level things to expect: first, the requirements.txt pulls in rembg and onnx for the segmentation nodes even though this clamp needs neither. Second, Allor's repo was rebased to drop all images from git history (a ~344x size cut), which can break auto-updates - the docs at nourepide.github.io/ComfyUI-Allor-Doc have dedicated troubleshooting pages for that. The ControlNet weights themselves are not bundled; load them as usual with ComfyUI's ControlNetLoader.

Real talk

ControlNetClamp is the most pointless-looking node in this pack and also one of the most honest: it does nothing, forever, on purpose. For the first few weeks of ComfyUI you should ignore it. The day you find yourself wiring one ControlNet into three apply nodes and squinting at the canvas, it'll click. Don't expect it to fix a misapplied ControlNet - no pass-through can - but it will make the graph that's confusing you actually readable.