📥 ControlNet Download

Paste a URL to a ControlNet model, name it, and this node downloads it and hands you a loaded CONTROL_NET - no manual folder-shuffling, no restart. It's the SDVN downloader wrapped around ComfyUI's ControlNet loader, aimed at the moment you find the exact depth or canny model you need and don't want to break flow to go install it.

What ControlNet does, and where this fits

ControlNet conditions your generation on spatial structure pulled from a reference image - edges, depth, pose, scribble - so the prompt decides what appears while the ControlNet decides where it goes. It was the single biggest control unlock in the model's history. To use it you need two things: a preprocessor that turns your reference into a hint map (a Canny edge image, a depth map), and the ControlNet model itself that reads that map. This node handles the second half - getting the model file - and it pairs with a preprocessor node (the SDVN pack recommends installing Fannovel16's controlnet_aux for those) and a ControlNet Apply node downstream.

The inputs and outputs that matter

• Download_url - a Civitai or Hugging Face link to the ControlNet model. Page address or direct link both work.
• Url_name - the filename to save as, e.g. control_v11p_sd15_canny.safetensors. Rename it from the default so you can tell your control models apart - you'll accumulate a lot of these.

Output is a single CONTROL_NET, which wires into a ControlNet Apply node alongside your conditioning and your preprocessed hint image.

Installing it

ComfyUI Manager: search SDVN_Comfy_node, install, restart. Or manually:

cd ComfyUI/custom_nodes
git clone https://github.com/StableDiffusionVN/SDVN_Comfy_node

Then pip install -r custom_nodes/SDVN_Comfy_node/requirements.txt from your ComfyUI root, and restart. Windows/macOS need aria2c installed for downloads to run. For the preprocessor side, the pack also recommends installing comfyui_controlnet_aux.

Where people get burned

Model has to match your base architecture. An SD 1.5 ControlNet will not work on an SDXL model, and neither works on Flux. Each base gets its own ControlNets. Modern bases mostly ship a single union model covering the standard set - canny, depth, pose, edge, scribble, gray, plus an inpaint mode - that lands within about a week of the base release, so for Flux or the 2025-26 models you're usually downloading one union checkpoint rather than a model per condition. Download the one built for your base.

The exotic conditions don't exist on new bases. If you came here for something like QR-code / brightness control, that's an SD 1.5 and SDXL-only relic - nobody rebuilt it for the newer architectures. New bases get the standard six or seven condition types and no more.

Preprocessor mismatch. A depth ControlNet needs a depth hint map; feed it a Canny image and you'll get a mess. The model and the preprocessor have to be the same type. This node only gets you the model - the hint map comes from your preprocessor.

Download failed. Missing aria2c, or a gated repo. Check the console for what it tried to pull.