CtrlNet ZoeDepth Pipe (JPS)

The other half of CtrlNet ZoeDepth Settings (JPS). That node bundles four ZoeDepth ControlNet values into one wire; this one splits the bundle back apart wherever downstream you actually need the individual numbers. Same trick as every Settings/Pipe pair in this pack - build the menu once, drop a thin unpack node wherever the graph needs it, instead of routing four separate wires across a busy canvas.

What comes out of it

One input, zoedepth_settings - a BASIC_PIPE (JPS's own bundle type, a plain tuple under the hood, unrelated to Impact Pack's object of the same type label). Four outputs:

• zoe_source (INT) - which reference image ZoeDepth reads from. It comes out as a plain integer rather than the "Source Image / Support Image / Support Direct" text, matching the select input on the pack's Image Switch node - the obvious wiring is straight into one of those.
• zoe_strength (FLOAT) - feeds your Apply ControlNet node's strength widget.
• zoe_start / zoe_end (FLOAT) - the step-window percentages for Apply ControlNet (Advanced)'s start/end inputs.

Notably leaner than its MiDaS sibling: CtrlNet MiDaS Pipe (JPS) has six outputs because MiDaS's preprocessor exposes two extra tuning parameters (midas_a, midas_bg). ZoeDepth doesn't get the same treatment here - there's nothing preprocessor-specific to unpack beyond source/strength/start/end, which tracks with ZoeDepth being a less commonly tuned option in this pack's ControlNet family to begin with.

This node runs no depth estimation and applies no ControlNet itself. It's pure routing between your settings menu and the nodes doing the actual work - a ZoeDepth preprocessor upstream turning your source image into a depth map, an Apply ControlNet node downstream conditioning generation on it.

Installing it

Comes with the pack. ComfyUI Manager: search JPS Custom Nodes for ComfyUI, install, restart. Manual route:

cd ComfyUI/custom_nodes
git clone https://github.com/JPS-GER/ComfyUI_JPS-Nodes.git

Restart afterward, and delete any prior install of the pack first per the README's own warning. Nothing to download for this specific node - the ZoeDepth model weights live with whichever preprocessor node actually runs the estimation, downloaded on first use.

Where people get tripped up

Without a CtrlNet ZoeDepth Settings (JPS) node feeding it, this node has nothing to unpack - it's not a standalone tool. Beyond that, the more useful thing to know is that ZoeDepth itself is a somewhat contested choice: it was built for metric (real-world-scale) depth, and the community's own comparisons against MiDaS have generally favored MiDaS's sharper edges for ControlNet conditioning specifically, even though ZoeDepth is the more geometrically accurate model on paper. If your depth-conditioned generations look soft around edges, that's worth investigating before assuming this pipe's wiring is wrong - the fix is likely a different preprocessor, not a different value flowing through this node. And as with the rest of this pack's ControlNet Settings/Pipe family, it's newer than the published README and won't turn up in that document's node list.