Anymatix ControlNet Loader

ControlNet is how you tell a diffusion model where things go instead of just what they are. Feed it a pose skeleton, a depth map, or edge lines as a condition, and the sampler keeps your character's stance or your scene's layout honest while the prompt handles the rest. AnymatixControlNetLoader is a path-taking twin of ComfyUI's stock ControlNetLoader: same job, same CONTROL_NET output, but the model name is a string rather than a dropdown.

Same rationale as every loader in this pack: Anymatix workflows address models by URL, so the ControlNet weights arrive via AnymatixFetcher and this node loads whatever path it's handed. The loader basenames the input and resolves it against models/controlnet, so a file already sitting there loads from just a filename too. It subclasses ComfyUI's own ControlNetLoader, which means it inherits the upstream loading behavior - this is a wrapper that changes where the name comes from, not a reimplementation that could drift from stock behavior.

Where it sits in a workflow

ControlNet only does something when you have a matching preprocessor. The usual chain: run AnymatixDWPreprocessor (pose), AnymatixHEDPreprocessor (soft edges), or a depth/edge node over your input image to get the condition map, load the ControlNet with this node, and feed both into Apply ControlNet alongside your model and positive conditioning. The strength dial lives on Apply ControlNet, not here - this node just loads the weights.

One genuinely useful thing to know from the wider ecosystem: the menu of ControlNet conditions has narrowed as the models have moved on. The pose, depth, and edge conditions that Anymatix ships preprocessors for are exactly the ones that survived the transition to newer base models - nobody rebuilt the old SD-era QR-code novelty, but pose and depth are everywhere.

Inputs and output

• control_net_name (STRING) - path or filename of the ControlNet weights.
• Output: CONTROL_NET → the control_net input on Apply ControlNet.

Install

Standard pack install:

cd ComfyUI/custom_nodes
git clone https://github.com/Anymatix/anymatix-comfy-nodes

restart, or use ComfyUI Manager (search "anymatix-comfy-nodes"). The node itself downloads nothing; ControlNet files are several GB and arrive via fetcher or manual download.

Issues to expect

The two mistakes that actually bite: version mismatch (a ControlNet trained for SDXL won't meaningfully condition an SD 1.5 model - and for the newer diffusion-transformers there are architecture-specific ControlNets, so "it loaded but does nothing" usually means the wrong family), and forgetting the preprocessor. If the ControlNet loads but your output ignores it entirely, check that a real condition map is actually reaching Apply ControlNet - a raw photo fed in as "the condition" is a different, much weaker thing than a clean pose or edge map.