ViT Age Classifier

This is the smallest node in the Runware pack and it's honest about it: one IMAGE in, one STRING out, nothing else. You feed it a picture of a person and it tells you the predicted age. That's the entire feature set, and for once the brevity is the point - there's not a single optional input to fiddle with.

The model is a ViT (Vision Transformer) trained as a classifier, and it runs on Runware's cloud as a caption task - the same machinery that powers image-captioning, here pointed at a single attribute. You get text back, not a number, so the output is a string like an age estimate. There's no widget config, no seed, no output format, no safety toggle. It's a utility node, and it behaves like one.

Why you'd reach for it

This node is not a creative tool; it's a guardrail and a data step. The obvious use: filtering. If you're generating people at scale, this is the cheap way to sanity-check that a batch didn't drift into generating children or octogenarians when your prompt said "adult." Wire it after an image-generation node and use the returned text in a switch or a check. The less obvious use: it's a labeler for dataset work - if you're curating a training set and need an age distribution, a cloud classifier that doesn't tax your GPU is a fine first pass.

It's also a useful reminder that this pack isn't only image and video generation. The same Runware catalog has text, audio, and these little one-job classifiers, and they all share the same install, the same API key, and the same pay-per-call billing. You're not paying for a model download you'll never use; you're paying for one classification call.

How it works

The node sends your image up as a caption-style request (runware:153@1) and converts the returned text into the text output. That output is a plain string, so it drops straight into any text-consuming node in your graph - a STRING socket, a condition, a file name. Because it's cloud inference, there's a network round-trip per call, and the node's title bar logs the run cost (which, for a classifier, should be the cheapest thing in your workflow).

Install

Same pack, same two ways:

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

Restart, and either search "Runware" in the Custom Nodes Manager or clone manually. You still need a Runware API key - Settings → "Runware API key", RUNWARE_API_KEY, or runware auth login - because even a one-job node is a paid API call.

The honest gotcha is expectations. A ViT age classifier is a broad-bucket predictor, not a legal ID check: it's decent at "child vs adult vs senior," shaky at 23 vs 27, and it guesses from appearance alone. Treat it as a coarse filter, not a measurement. And because it's cloud-based, there's no offline mode - the node does nothing until it can reach Runware. For a node this thin, that's the only real caveat: it's a nice toy, but it's a cloud toy.