debugInput

Every ComfyUI user eventually hits the moment where a workflow "works" but the output is wrong, and you can't tell whether it's the prompt, the seed, or the conditioning that's quietly doing something dumb. debugInput from the tinyterraNodes pack is the node you reach for then: you attach it to any wire and it shows you the actual value sitting on it, right inside the node.

It's an output/debug node, so it has no outputs at all - think of it as a terminal you clip onto the graph. The clever bit is the debug input, which is typed as * (any type). That means it will happily swallow an IMAGE, a LATENT, conditioning, a mask, text, or a number. Whatever you connect, the node stringifies it and renders the text in its body. A LATENT prints a dict with a samples tensor; conditioning prints a list of tensor + metadata pairs; plain strings print as themselves. Most of the time that's enough to spot the problem.

The inputs that matter

• debug - the wire you're inspecting. Anything goes.
• print_to_console - off by default. Flip it on and the same value gets dumped to ComfyUI's terminal instead of (or in addition to) the node body.
• console_title - a label prefixed to the console output, default ttN debug:. Useful when you've got three debug nodes and can't tell whose line is whose.
• console_color - 16 colors (Red, Green, Cyan, LightWhite, etc.) so you can color-code different debug points. Purely cosmetic.

Because it's marked an output node, anything upstream of it executes even if you don't need it for the final image - which is exactly why you want it while debugging and don't want to leave a pile of them wired into a finished workflow. It forces work that can eat VRAM and slow a run for nothing.

Installing it

debugInput ships inside the tinyterraNodes pack, so it's one install for the whole family. In ComfyUI Manager: Install Custom Nodes → search "tinyterraNodes" → Install → restart. Or from a terminal:

cd ComfyUI/custom_nodes
git clone https://github.com/TinyTerra/ComfyUI_tinyterraNodes

No model downloads and no pip dependencies - this node is pure Python plus the pack's front-end JS. After cloning, restart ComfyUI and the node appears under 🌏 tinyterra/util.

Gotchas

Two things trip people up. First, the node shows a textual view: an IMAGE input prints as a tensor shape like torch.Size([1, 512, 512, 3]), not a picture. If you want to actually see the pixels, use the pack's imageOutput node set to Preview instead - debugInput tells you data exists, not what it looks like. Second, remember it's a dead-end: the wire still flows from your source node, so you can attach debugInput without re-routing anything, but the node itself consumes nothing and passes nothing on. That's the whole job - look, learn, delete.