Text-to-video generation on VBench (VBench and Degradation Metrics)
79.13VBench ScoreOpenSora
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| OpenSoraSampling steps=30 steps, Latency (s)=102.287, FLOPs (T)=3283.20, Speedup=1.00×2026.03 | 79.13 | — | |
| SODANumber of samples (Ns)=21, Latency (s)=86.684, FLOPs (T)=2319.94, Speedup=1.42×2026.03 | 79.13 | 0 | |
| SODANumber of samples (Ns)=12, Latency (s)=49.896, FLOPs (T)=1314.42, Speedup=2.50×2026.03 | 78.49 | 8.08 | |
| SODANumber of samples (Ns)=11, Latency (s)=48.022, FLOPs (T)=1277.59, Speedup=2.57×2026.03 | 78.41 | 9.09 | |
| DuCaLatency (s)=50.637, FLOPs (T)=1315.62, Speedup=2.50×2026.03 | 78.39 | 9.35 | |
| ToCaLatency (s)=54.698, FLOPs (T)=1394.03, Speedup=2.36×2026.03 | 78.34 | 9.98 | |
| ∆-DiTFLOPs (T)=3166.47, Speedup=1.04×2026.03 | 78.21 | 11.62 | |
| PABFLOPs (T)=2558.25, Speedup=1.28×2026.03 | 76.95 | 27.55 | |
| FORAFLOPs (T)=1751.32, Speedup=1.87×2026.03 | 76.91 | 28.06 |