
10,312 hours of unscripted egocentric household video under Apache 2.0, split into raw and captioned tiers, provides the kind of first-person manipulation data that world models and vision-language-action work requires and that no one uploads voluntarily.
Checking sign-in…
Loading comments…