Song-Aware Planning, Generation, and Assembly for AI Music Videos
A generative video model can make a good four-second clip. A music video takes forty of them, and they have to land on the right beat, sing the right words, cut on purpose, and keep the same face from shot to shot. Prompting clips one at a time and dragging them into an editor falls apart at the second revision. Storyboard Director treats the film plan as the source of truth. It analyzes the whole song, drafts a shot-by-shot plan with a local language model, and gates expensive rendering behind an explicit approval. Every render goes back onto one shared music timeline against the untouched master audio.