Research
Berkeley researchers predict egocentric video from full-body human actions
UC Berkeley's PEVA model predicts future egocentric video frames conditioned on 3D pose changes, enabling long-horizon video generation and counterfactual simulation for embodied AI agents.
1 min read
SourceBerkeley AI Research
UC Berkeley researchers released PEVA (Predicting Ego-centric Video from human Actions), a world model that generates future egocentric video frames conditioned on full-body human motion. The model takes past video frames and a desired 3D pose change as input, then predicts the next video frame, ena...
Sign in to read the full analysis
Free account. Full analysis on LLM unit economics, plus the weekly Cost-of-Inference column.
Method & sources
- Source type
- Primary publication (lab/vendor blog) — our analysis + implication
- Source link
- Berkeley AI Research
- Published
- UTC
- Byline
- By the gotcontext.ai team (editorial standards)
- Correction?
- corrections@gotcontext.ai