DataMagic: Authoring Data Videos through Declarative Multi-Agent Orchestration
DataMagic turns raw tables into narrated data videos by using a declarative spec plus coordinated agents.
The paper says DVSpec represents charts, narration, animations, timing, and data-bound references in one shared format. Its “Generate-then-Orchestrate” setup creates candidate scenes in parallel, then optimizes them for narrative coherence. In evaluations on 109 samples, DataMagic scored 3.89/5 versus 2.13/5 for GPT-5, with execution success above 95%. A user study found a 79.7% reduction in task time compared with a conversational LLM workflow. HF Daily Papers' note
The paper says DVSpec represents charts, narration, animations, timing, and data-bound references in one shared format. Its “Generate-then-Orchestrate” setup creates candidate scenes in parallel, then optimizes them for narrative coherence. In evaluations on 109 samples, DataMagic scored 3.89/5 versus 2.13/5 for GPT-5, with execution success above 95%. A user study found a 79.7% reduction in task time compared with a conversational LLM workflow. HF Daily Papers' note
score 4