World Labs' RTFM generates new views of a scene from images, allowing someone to explore it through a stream of generated video. The research announcement describes a model that learns to render those views without first constructing an explicit geometric model of the scene.
Its memory is organized by where each previous frame was viewed from. When generating another view, the system retrieves nearby frames instead of consulting the entire growing history. World Labs calls this context windowThe maximum amount of text a model can consider at once — its working memory for the current conversation or task.Full definition → juggling: it is the mechanism intended to preserve previously visited places while keeping each rendering step manageable.
Terms in this piece · Glossary
context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.
