Phantom

Phantom: Subject-Consistent Video Generation via Cross-Modal Alignment