ResearchBenchmarksText-to-Video Generation on Wan 1.3B Text-to-Video 480P 2.1Follow97.36Temporal FidelityDense96.756896.913497.0797.2266Apr 14, 2026Evaluation ResultsMethodMethodLinksTemporal FidelityMotion ScoreAesthetic QualitySSIMPSNRLPIPSLatency (s)Speedup (×)DenseSparsity=0.00%Sparsity=0.00%2026.0497.3698.2766.34———1211PASASparsity=85%Sparsity=85%2026.0497.0698.1265.3375.120.510.2028711.7PISASparsity=85%Sparsity=85%2026.049798.0865.1772.8619.660.2246711.7SVG2Sparsity=84.5%Sparsity=84.5%2026.0496.7898.0465.5277.6521.520.1598861.4