DAViD Depth Visualizer

A raw depth map is just numbers

A depth map out of the model is a single-channel grayscale image: closer is brighter, farther is darker, and honestly it's hard to read at a glance. DAViDDepthVisualizer is the utility that turns that into something you'd actually show someone - it applies a colormap of your choosing, with controls for range and orientation. It's a small node, and that's fine; it does one job and does it cleanly.

What it does mechanically

It takes a depth_raw IMAGE, optionally negates the values (invert), normalizes the range, then runs it through OpenCV's colormap function. If you feed it a three-channel image it converts to grayscale first, so it won't crash on a stray colored input - it'll just quietly desaturate it. Output is a single depth_visualization IMAGE, same batch size as what went in.

The settings you'll actually touch

• colormap - the one you'll change. Nine options: TURBO is the default and the one the pack's own multi-task node uses (readable, high contrast); JET is the old classic everyone recognizes; GRAY gives you a clean single-channel height map you can genuinely use; the rest (HSV, VIRIDIS, PLASMA, INFERNO, MAGMA, BONE) are Matplotlib refugees for when you want a specific look.
• min_depth / max_depth - set these with normalize on to lock the mapping range instead of letting values auto-stretch. Default 0 to 1.
• invert - flips near/far. Some pipelines want white = far.
• normalize - leave it on unless you know your depth scale and want raw values.

Where the raw depth comes from - read this

Here's the honest part: this pack's own DAViDMultiTask node outputs its depth already TURBO-colored, so this visualizer's natural input is single-channel raw depth - the pack's individual per-task depth models, or a depth map from any other source (Depth Anything, Marigold, comfyui_controlnet_aux preprocessors). Feed it the multi-task node's colored output anyway and it'll desaturate and recolor it, which is a perfectly good way to switch colormaps on the same data. It's a genuinely thin utility node, and that's okay - grab it when you want a specific colormap or a printable grayscale version of your depth.

Same pack, same install

This node ships in AIWarper's ComfyUI-DAViD pack alongside the multi-task node, so the install is the shared one:

cd ComfyUI/custom_nodes
git clone https://github.com/AIWarper/ComfyUI-DAViD
pip install onnxruntime-gpu

It needs no model of its own - it's pure post-processing - but the pack's models live in ComfyUI/models/david if you're using the multi-task node too. Restart ComfyUI and it's under DAViD/Utils.

Troubleshooting

If the output is one solid slab of color, your min_depth/max_depth range is degenerate (min ≈ max) or you turned normalize off on values that fall outside 0–1 and everything clipped. Reset the range or flip normalize back on. For best results feed it a true single-channel map rather than an already-colored one.