Google Omni turns map scribbles into virtual drone flights via multimodal video

A viral post on X showcased Google's Omni model, a multimodal video AI that can interpret a hand-drawn red line on a Google Earth screenshot as a 3D camera path. The model then generates a smooth, first-person drone flight video following that path, effectively creating a virtual camera presence where none physically exists. This goes beyond simple image analysis—Omni understands spatial intent from a crude scribble and synthesizes a coherent video sequence with realistic motion, perspective changes, and scene transitions. The demo highlights Omni's ability to combine visual grounding, video generation, and path planning into a single operation, a significant advance over prior models that required explicit waypoints or pre-recorded tracks.

For professionals, Omni's scribble-to-video capability opens up instant virtual tours for real estate, geospatial surveys, and urban planning—without needing a drone or on-site camera. A realtor could sketch a path on a satellite image and get a sellable flythrough; an urban planner could visualize proposed routes through a neighborhood. It also hints at future applications in film pre-visualization, gaming, and remote exploration. As multimodal models like Omni grow more adept at translating abstract human input (a squiggly line) into rich, dynamic media, the barrier between idea and production collapses—democratizing video creation for anyone with a map and a marker.