MapStory: Prototyping Editable Map Animations with LLM Agents
summary
MapStory generates editable map-animation prototypes from plain text using a dual-agent LLM architecture that first produces a scene breakdown of camera moves, highlights, and animated elements, then a researcher agent with web search automatically retrieves GeoJSON regions, paths, and coordinates. Users edit timeline items and tweak primitive parameters; demos include a Nigeria highlight, a ship route for a ceremonial mace, and historical Andhra Pradesh boundaries. Design draws on an analysis of 200 map videos and evaluation with N=5 experts and N=12 users.
AI-written summary · about
chapters (8)
- 0:00Introduction and motivation
- 1:55Map animations today
- 2:59Live demo of MapStory
- 4:25Complex scene breakdown
- 6:35Expert insights and research agent
- 9:26Animation primitives and system
- 11:34Researcher agent demo
- 12:37Closing and Q&A
related talks
25mMapping diplomacy: Visual storytelling for global impact
45mBuild an agent with Mapbox MCP
01h 34mBuild a Web Map with Open Source Tools
50mAdvanced map style customization in Mapbox Studio
25mArtificial intelligence, meet location intelligence
37mSlippy Map Style: Geo-canvas design with CartoCSS
12m"Look! A Map"
33mMaps on the Web
