Video Depth Anything Output Depth Map

VideoDepthAnythingProcess hands you back a DEPTHS object, and that's not something you can preview, save, or feed to a ControlNet node - it's this pack's internal, full-precision container for raw depth data. This node is the converter: it turns that opaque object into a plain IMAGE, the type every other node in ComfyUI already knows how to deal with. Small node, but it's the one that makes the rest of the pipeline usable outside itself.

The one choice you make: colormap

There's really one decision here - colormap, a dropdown of gray (the default) or inferno:

• gray is the standard depth-map convention: white is close to the camera, black is far, everything else a smooth gradient between. This is what you want if the output is heading anywhere near a ControlNet apply node, or into any pipeline that expects a conventional depth map - it's the format the rest of the ecosystem was built around.
• inferno applies a perceptual heatmap palette instead - the same kind of colorful, high-contrast gradient you'd see on a thermal camera or a matplotlib depth plot. It's genuinely easier to read at a glance, since human eyes pick out color-coded distance bands faster than shades of gray, especially in a busy scene. Use it for a human-facing preview or a stylized visual, not for anything downstream that expects a real depth map - ControlNet and similar conditioning nodes are trained on grayscale depth and won't interpret a colorized heatmap correctly.

The honest way to think about it: gray is the working format, inferno is the presentation format. If you're not sure which you need, you need gray.

Output

One output, images (type IMAGE) - a completely ordinary batch of images, one per frame. From here it behaves like any other image batch: run it through a Preview Image node to look at it, a Save Image node for stills, or a video-combine node (from whichever video pack you're already using) to reassemble it into an actual depth-map video file.

Installing it

ComfyUI Manager: search ComfyUI Video Depth Anything, install, restart. Or manually:

cd ComfyUI/custom_nodes
git clone https://github.com/yuvraj108c/ComfyUI-Video-Depth-Anything

then restart ComfyUI. This node is only useful once you've got LoadVideoDepthAnythingModel and VideoDepthAnythingProcess upstream feeding it a real DEPTHS object - it's not independently useful, and there's nothing extra to download for this node specifically beyond installing the pack itself.

Common issues & troubleshooting

Colors look wrong when fed into ControlNet. This is almost always the inferno colormap picked by mistake. ControlNet's depth conditioning expects a grayscale map where pixel brightness alone encodes distance - feed it a colorized heatmap and the conditioning signal it reads is nonsense. Switch back to gray.

Nothing shows up / node errors immediately. Check that depths is actually connected to a VideoDepthAnythingProcess output and that node ran successfully first - this node has nothing to convert if the upstream depth pass didn't complete. It's a pure converter with no model of its own, so a failure here almost always traces back to the node before it.

Output looks flat or low-contrast even on gray. That's more likely a VideoDepthAnythingProcess or model-choice issue than something to fix here - check whether you accidentally loaded one of the metric_ checkpoints on LoadVideoDepthAnythingModel, since those are tuned for absolute distance rather than the sharp relative contrast that makes a depth map easy to read.