Video Reconstruction on TokenBench
36.12PSNRVidTok
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| VidTokParameters=157M, Latent Channels=162025.03 | 36.12 | 0.9511 | 0.0166 | 40.89 | |
| LeanVAEParameters=40M, Latent Channels=162025.03 | 35.71 | 0.9506 | 0.0173 | 44.63 | |
| WF-VAEParameters=316M, Latent Channels=162025.03 | 35.11 | 0.9435 | 0.0222 | 48.03 | |
| CogVideoX-VAEParameters=206M, Latent Channels=162025.03 | 34.39 | 0.9366 | 0.0331 | 75.54 | |
| Cosmos-TokenizerParameters=101M, Latent Channels=162025.03 | 33.13 | 0.9191 | 0.0669 | 109.27 | |
| VidTokParameters=157M, Latent Channels=42025.03 | 31.38 | 0.8963 | 0.0526 | 178.8 | |
| Open-Sora VAEParameters=393M, Latent Channels=42025.03 | 31.27 | 0.8883 | 0.0774 | 282.81 | |
| LeanVAEParameters=40M, Latent Channels=42025.03 | 31.12 | 0.8957 | 0.0432 | 162.55 | |
| OD-VAEParameters=239M, Latent Channels=42025.03 | 30.47 | 0.8791 | 0.0618 | 211.93 | |
| CV-VAEParameters=182M, Latent Channels=42025.03 | 30.37 | 0.8758 | 0.0706 | 265.99 |