A model that learns actions from video and turns them into action instructions.
A Video-Action Model is like copying a TikTok dance in the kitchen. It watches the move, then says, “Left foot, please. Not the fridge.”
It helps robots and game characters. It also helps video agents turn video into action.
Embodied Video Model
Both use video as action experience, but focus on different parts.
VLA
It can join with vision and language to control actions.
Video-to-Robot Learning
It helps move from video demos to robot actions.
Embodied AI
It helps Embodied AI not just see the world, but act in it.