ResearchBenchmarksSubject-driven Video Generation on Subject-to-Video (S2V) (test)Follow0.538MSLibraGen0.37680.418650.46050.50235Mar 13, 2026Evaluation ResultsMethodMethodLinksMSMQAESIQAVQTALibraGen2026.030.5380.9930.649671.62.7953.594Vidu Q12026.030.53730.99240.649171.822.7953.315Kling O12026.030.49650.98650.647972.842.7973.567Phantom2026.030.38440.98730.64169.352.3731.998MAGREF2026.030.3830.98530.635670.362.3541.784