Handshape practice · graded on your device

The mirror that knows the alphabet.

Watching a letter is the easy half. Here you make it: the avatar forms the target, your camera watches your hand, and the page scores the handshape live, finger by finger. Hold it right and the letter passes. Hold it wrong and it names the finger. Nothing your camera sees ever leaves this tab.

0 of 26 letters held correctly

The target

watch, then copy

Target shape

Closed fist, thumb resting straight along the side of the index finger.

Your hand

graded in this tab only

The camera is off. Turn it on to be graded live. The video, the hand landmarks, and every score stay in this tab: there is no network call in the practice loop. Or skip the camera and copy the target diagram by eye.

Downloading the hand tracker. About 8 MB, once per browser. It is cached after this, and it runs in this tab from then on.

Waiting for the camera. Allow the camera in your browser prompt and the grading starts.

Turn the camera on to be graded, or just copy the diagram.

The alphabet, in the order it is easiest to learn

Closed fists first, open hands next, then the pairs every beginner confuses. A letter turns green once you have held it correctly; your progress is kept on this device. You can also press any letter key to jump straight to it.

This is the practice half of the signing loop. The reference for every letter and number lives at /asl-alphabet, whole words and signed chat replies at /sign-language, and the engine behind all three pages is documented in the sign language docs and shipped as the @axis/sign-language package, installable from a clone until its registry publish lands. The avatar, the target diagram, and the grader all read the same handshape spec, so what is taught and what is graded can never drift apart.