A new preprint on arXiv introduces MOMO, a model for generating or modeling human motion from moving sounds. The abstract argues that our responses to sound are shaped not only by what we hear, but by the evolving spatial relationship between us and the source. That relationship is carried by spatial audio and its changes over time.
According to the abstract, MOMO treats those temporal dynamics as central to understanding human reactions. The paper appears to use the motion of a sound source—not just its location at a single moment—as the input for producing corresponding human movement. However, the available excerpt is brief, so the exact architecture and evaluation details are not yet described.
The framing points to a broader idea: spatial audio is not merely a realism feature, but an informative signal about the world that shapes physical response. MOMO is one attempt to make that link explicit, though whether it targets subtle gestures, full-body movement, or both remains unclear from the abstract alone.