Hand Wave

Hand Wave reads hand shapes straight from your webcam. It tracks landmarks on both hands and turns them into recognized signs as you move, so you see output the moment you finish a gesture. The hand gesture recognition here is frame by frame. No upload step, no waiting on a file.

You can feed the tool a shared screen instead of a camera. That lets you point it at a video call, a stream, or a playback of a recording and still get recognition. Handy. It's the practical choice when the person signing isn't sitting in front of your own webcam.

Everything runs in a web page. There's no app store download, no desktop install, and no account wall before you start. Open the site, grant access to a camera or screen, and recognition begins on its own. It works on Windows, macOS, and Linux Chrome-based browsers, which makes it one of the more approachable accessibility tools for a quick session.

Frames stream over a WebSocket to a remote inference endpoint, which returns results frame by frame rather than in one big batch at the end. The app exposes its own timing readouts, including hand frames per second and round-trip latency, so you can tell when a slow connection, not the signing itself, is the reason a sign gets missed. Watch the numbers.

The tool tracks left and right hands separately, which matters since many signs depend on the relationship between both hands. When only one hand is in frame, recognition narrows to that hand. Add the second hand back and you get a fuller read.

A trace view shows hand detection counts and per-stage timing. If recognition feels sluggish, you can check whether the delay lives in detection, inference, or the network round trip instead of guessing. It's a small feature. It saves a lot of guesswork when something looks off.