| video-generation-on-kinetics-600-12-frames-64x64 | MAGVIT-v2 | #1 | FVD: 4.3±0.1 |
| video-prediction-on-kinetics-600-12-frames-64x64 | MAGVIT-v2 | #1 | FVD: 4.3±0.1 |
| video-generation-on-ucf-101 | MAGVIT-v2 | #4 | FVD16: 58±3 |
| video-generation-on-ucf-101 | MAGVIT-v2 (AR) | #8 | FVD16: 109 |
| image-generation-on-imagenet-512x512 | MAGVIT-v2 | #34 | FID: 1.91Inception score: 324.3 |
| image-generation-on-imagenet-512x512 | MAGVIT-v2 (w/o guidance) | #61 | FID: 3.07Inception score: 213.1 |
| action-recognition-in-videos-on-something-something-v2 | MAGVIT-v2 + ViT-B (tokens as targets) (extra data) | #59 | Top-1 Accuracy: 67.38 |
| image-generation-on-imagenet-256x256 | MAGVIT-v2 | #70 | FID: 1.78 |
| image-generation-on-imagenet-256x256 | MAGVIT-v2 (w/o guidance) | #166 | FID: 3.65 |