MAGI 2 Showcases
Made with MAGI 2
Capabilities of the unified audio-video generation model - from lip-synced character performance to rhythm-aware music-to-video.
Character performance
Dialogue, singing and emotional shifts with natural lip-sync, facial expressions and body language.
Music to video
Turn a track into synchronized visuals with rhythm-aware camera moves and scene cuts.
Long-form scenes
One token sequence keeps story, motion, sound and camera rhythm coherent across longer generations.
MagiMoE efficient scaling
114B parameters with only about 6B activated per token, so long video stays practical to generate.