Execute All ControlNet Preprocessors

This is the node you run when you don't yet know which preprocessor you want. Feed it one image and it runs the whole preprocessor lineup in this pack - canny, depth, softedge, lineart, pose, normals, segmentation, the lot - and hands you back a single image so you can eyeball all the outputs side by side. It's a testing and exploration tool, not something you leave wired into a production graph. Think of it as a contact sheet for ControlNet hints.

What it's actually for

Picking a preprocessor is half the battle in ControlNet work, and the honest way to pick is to look. A canny map of a soft, organic subject looks like static; a HED softedge of a hard mechanical object throws away the crisp lines you wanted. Rather than swap a preprocessor node twenty times and re-run, you run this once and see every result at a glance. The pack even ships this as its "testing workflow" - the ExecuteAll.png example in the repo is literally this node with a demo image.

Once you've spotted the winner, you don't keep this node around. You drop in that one dedicated preprocessor node (say DepthAnythingV2Preprocessor or CannyEdgePreprocessor), wire it into a ControlNet Apply node with the matching model, and move on. This node's job is done the moment you've made the decision.

The inputs and output

There are only two inputs, and one of them you'll rarely touch:

• image - the picture you want to see processed every possible way.
• resolution (default 512) - the working size each preprocessor runs at. Match it roughly to your render resolution; leave it at 512 if you're just scouting.

The single output is an IMAGE containing all the preprocessor results. You typically wire it straight into a Preview Image or Save Image node and study it. Don't bother feeding it into a ControlNet Apply - it's a composite of many different hint types, not a usable conditioning map on its own.

Installing it

ComfyUI Manager is the easy path: search ComfyUI's ControlNet Auxiliary Preprocessors, install, restart. Manual works too:

cd ComfyUI/custom_nodes
git clone https://github.com/Fannovel16/comfyui_controlnet_aux
pip install -r requirements.txt

Then restart ComfyUI.

Where people get burned

The big one is model downloads. Because this node runs every preprocessor, the first execution wants every model file - depth weights, HED, pose detectors, segmentation checkpoints, normal estimators, and more, all pulled from HuggingFace on demand. That first run can be a long, bandwidth-heavy pause and eats a chunk of disk. That's expected, not a hang; let it finish. If you're on a metered connection or tight on storage, run the individual preprocessors you actually care about instead.

Second: it's slow by design. You're running a dozen-plus models on one image. Don't put it in a batch loop or a video pipeline - it's a one-shot scouting tool.

Third, the universal ControlNet reminder that catches everyone eventually: a preprocessor only makes the hint image. Nothing here conditions your generation until you load the matching ControlNet model and route a real hint through a ControlNet Apply node. This node just tells you which hint to go make properly.