Where the numbers come from.

Every claim we make should be checkable. The references below back up the statistics and the legal context we cite across the site. If you find an error or a stronger source, tell us — we'll update it.

"7.5 million Americans have trouble using their voices"

Approximately 7.5 million people in the United States have trouble using their voices. Source: the National Institute on Deafness and Other Communication Disorders (NIDCD), Statistics on Voice, Speech, and Language. This is the figure we use whenever we say voice-only design excludes a group much larger than people expect — and it counts only voice difficulty, not the millions more who are Deaf or hard of hearing.

"~5 billion smartphones, all with capable cameras"

Roughly 4.9–5 billion smartphones are in active use worldwide, per recent industry reporting (Statista and GSMA Intelligence). Every current smartphone includes a camera and enough on-device processing to run hand-landmark recognition locally. The point of the number is simple: the hardware our interfaces need is already in nearly everyone's pocket.

Drive-thru access and ADA litigation

Drive-thru ordering is a recurring subject of ADA-related advocacy and litigation for Deaf and hard-of-hearing customers. Notable matters include Magee v. McDonald's USA, LLC (E.D. La.), resolved through settlement in 2018, and a comparable action involving Burger King in 2019; the National Association of the Deaf has advocated on the issue for years. We cite these as context, not legal advice — the takeaway is that the access gap is real and already contested.

"Learnable in an afternoon"

When we say the 36 handshapes for A–Z and 0–9 are learnable in an afternoon, that's an estimate consistent with introductory ASL curricula used in Gallaudet University outreach and elementary-school exposure programs. It is a practical estimate, not a peer-reviewed measure — fluency is a lifelong practice; recognizing and forming the static alphabet is the part most people pick up quickly.

The recognition pipeline

Hand detection runs on an open-source, on-device 21-point hand-landmark model — the same class of pipeline used widely in research labs and consumer apps. It compiles and runs in the browser; camera frames are processed locally and never leave the device — only the parsed token (a letter, a digit, a gesture) reaches anything downstream. The handshape classifier and the context-aware dispatch built on top of that model are 1891's own work. For a research-grade vocabulary, custom training would be the appropriate next step.

Every claim on the site should be checkable. If you find a stronger source, that's a contribution — send it.

If you cite us, please cite 1891 Immersive Experiences, Frederick, MD and link the page you're drawing from. If we cite you, we'll do the same.