VocaLoci, a speech assistant that offers the word you were reaching for

Why it surfaced

This is a focused accessibility product with unusually concrete evidence: live-session latency of 491–508 milliseconds, top-three prediction accuracy, and a 58-second recorded demo. Its early pilot figures are explicitly labeled directional rather than definitive, giving the waitlist page more candor than most product launches.

VocaLoci listens as you speak and suggests likely words in real time, usually within about half a second, while keeping processing on your device by default.