NVIDIA's Signs turned every webcam into an ASL tutor that feeds its own dataset

Around 90% of deaf children are born to hearing parents, and traditional ASL classes are expensive and geographically uneven; video tutorials exist but cannot tell a learner their hand position is wrong. That missing feedback is why self-study of sign language so often fails.

NVIDIA, Hello Monday/DEPT and the American Society of Deaf Children launched Signs in February 2025: a free web platform where a 3D avatar demonstrates each sign and AI tracks the learner's hands through any webcam, giving instant corrections. At launch it offered 100 validated signs, with a goal of 1,000.

Learners' own videos become contributions to an open-source ASL dataset that trains better recognition models, making the resource more valuable with every user. Launched with zero paid media, Signs was covered by CNN, Axios and VentureBeat, and won a Silver Pencil at The One Show 2025.

Every learner's practice improves the AI dataset

Real-time correction fixes the fatal gap in video learning

Learners double as dataset contributors for free

An open dataset positions NVIDIA in accessibility AI

Zero paid media meant the tool had to earn its reach

The missing ingredient in a learning market is feedback: make the product itself the teacher that watches you, and let every practice session quietly build the asset that makes it smarter.

NVIDIA said it aimed to grow the platform from 100 validated signs at launch toward 1,000, and the learner-built, open-source dataset continues to expand for sign-language research. The One Show awarded Signs a Silver Pencil in Innovation in AI in 2025.