[Deprecated] ControlNet Preprocessor

Let's lead with the honest part: this node is deprecated, and the author says so himself. It was a convenience wrapper that let you switch between a bunch of ControlNet preprocessors (canny, depth, openpose, lineart, and friends) from a single node with a dropdown, instead of dragging in a separate node for each one. Nice idea. But the thing it wrapped - the comfyui_controlnet_aux pack - now ships its own all-in-one node, AIO_Preprocessor, which does exactly this. So Art Venture's version is on the way out, and the recommendation from the pack itself is: use AIO_Preprocessor directly.

If you landed here because a workflow uses AV_ControlNetPreprocessor, this article is mostly to explain what it did and steer you to the replacement. It's the highest-searched node in the pack, and almost every one of those searches is somebody hitting the "why does this only show None?" wall (more on that below).

What it did and how it worked

ControlNet needs a preprocessed version of your image - a depth map, an edge map, a pose skeleton - as its conditioning input, not the raw photo. This node was a router: pick a preprocessor from the dropdown, and it ran that specific detector from comfyui_controlnet_aux and handed back the processed map, plus the matching ControlNet model name so you could load the right control model. One node, many preprocessors, less graph clutter. The catch is baked into the design - it doesn't contain any preprocessors. It's a front-end that calls into comfyui_controlnet_aux, and if that pack isn't installed, the node has nothing to offer.

The inputs and outputs

• image - the image to preprocess.
• preprocessor - the dropdown of available preprocessors. Here's the tell: if this only shows None, comfyui_controlnet_aux isn't installed, so the list is empty. That's the single most common problem with this node.
• sd_version - sd15 or sdxl, so it picks a matching control model name.
• resolution (optional) - the preprocessing resolution, default 512, 64–2048 in steps of 64. Match it to your generation resolution for best alignment.
• preprocessor_override (optional) - a string to force a specific preprocessor by name.

Outputs: the preprocessed IMAGE (wire into a ControlNet Apply node) and CNET_NAME, a STRING naming the suggested control model.

How to install it

It's in the Art Venture pack (ComfyUI Manager → search comfyui-art-venture → Install → restart; or git clone https://github.com/sipherxyz/comfyui-art-venture into ComfyUI/custom_nodes). But installing Art Venture alone is not enough for this node to work - read on.

Common issues & troubleshooting

The preprocessor dropdown only shows "None." This is the issue, and it trips up newcomers constantly. The node is a wrapper around comfyui_controlnet_aux, so if that pack isn't installed, there are no preprocessors to list. The fix people land on: install comfyui_controlnet_aux via ComfyUI Manager (search it, install, restart). Once it's present, the dropdown fills up.

Should I even use this? Probably not for new work. The pack has deprecated it in favour of AIO_Preprocessor from comfyui_controlnet_aux - same all-in-one convenience, actively maintained, no wrapper layer. If you're building fresh, drop in AIO_Preprocessor and skip this entirely. Keep AV_ControlNetPreprocessor only to keep an old workflow running.

Worth stepping back: in 2026, instruction-editing models will take a preprocessed control map plus a sentence and follow it without a ControlNet checkpoint at all. You still need the map (that's what a preprocessor makes), but the downstream plumbing is getting simpler. Either way, make your maps with AIO_Preprocessor, not this.