Video Reconstruction on UCF-101
3.11rFVDHYDRA-XTOK (Stage 3)
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| HYDRA-XTOK (Stage 3)Compression Spatial (fs)=16×, Compression Temporal (ft)=4×, Input Resolution=256×2562026.06 | 3.11 | 36.88 | — | 0.967 | — | |
| Wan2.2Compression Spatial (fs)=16×, Compression Temporal (ft)=4×, Input Resolution=256×2562026.06 | 4.15 | 36.11 | — | 0.961 | — | |
| AToken-So/C (Stage 3)Compression Spatial (fs)=16×, Compression Temporal (ft)=4×, Input Resolution=256×2562026.06 | 7.77 | 34.66 | — | 0.953 | — | |
| MAGVIT-v2-MLMGenerator #Params=307M2026.03 | 8.6 | — | — | — | 1,280 | |
| MAGVIT-v2-ARGenerator #Params=840M2026.03 | 8.6 | — | — | — | 1,280 | |
| EVATokTokenizer #Params=145M, Generator #Params=327M2026.03 | 9.7 | — | — | — | 774 | |
| EVATokTokenizer #Params=145M, Generator #Params=633M2026.03 | 9.7 | — | — | — | 774 | |
| TivTok-T512#Tokens=512, #Dim.=32, T/P (%)=0.0492026.06 | 12.65 | 30.26 | 0.0533 | 0.8982 | — | |
| Cosmos-M#Tokens=2048, #Dim.=16, T/P (%)=0.1952026.06 | 13.67 | 31.7 | 0.0575 | 0.9177 | — | |
| Omni#Tokens=4096, #Dim.=8, T/P (%)=0.3912026.06 | 14.53 | 29.34 | 0.0487 | 0.925 | — | |
| TivTok-T1024#Tokens=1024, #Dim.=16, T/P (%)=0.0982026.06 | 17.97 | 29.54 | 0.0607 | 0.8897 | — | |
| LARP#Tokens=1024, Codebook Size=8192, Resolution=128x128, Frames=16, Extra Training=additional 75 epochs2026.01 | 20 | 28.71 | 0.075 | — | — | |
| LARP-L-LongTokenizer #Params=173M, Generator #Params=343M2026.03 | 20 | — | — | — | 1,024 | |
| LARP-L-LongTokenizer #Params=173M, Generator #Params=632M2026.03 | 20 | — | — | — | 1,024 | |
| TivTok-T128#Tokens=128, #Dim.=128, T/P (%)=0.0122026.06 | 21.29 | 30.13 | 0.0614 | 0.901 | — | |
| STAT-Video#Tokens=1297, Codebook Size=8192, Resolution=128x128, Frames=16, Extra Tokens=3002026.01 | 22 | 29.4 | 0.059 | — | — | |
| STAT-Video#Tokens=1197, Codebook Size=8192, Resolution=128x128, Frames=16, Extra Tokens=2002026.01 | 23 | 29.27 | 0.06 | — | — | |
| TivTok-T512Video resolution=32×256×256, #Tokens=640, #Dim.=322026.06 | 23.26 | 30.25 | 0.0591 | 0.8948 | — | |
| SDXL-VAE#Tokens=16384, #Dim.=4, T/P (%)=1.5632026.06 | 23.68 | — | — | — | — | |
| LARP#Tokens=1024, #Dim.=16, T/P (%)=0.391, Video resolution=16x128x1282026.06 | 23.93 | 28.65 | 0.0425 | 0.9003 | — | |
| LARP#Tokens=1024, Codebook Size=8192, Resolution=128x128, Frames=162026.01 | 24 | — | — | — | — | |
| MAGVIT-MLMTokenizer #Params=158M, Generator #Params=306M2026.03 | 25 | — | — | — | 1,024 | |
| MAGVIT-ARTokenizer #Params=158M, Generator #Params=306M2026.03 | 25 | — | — | — | 1,024 | |
| STAT-Video#Tokens=1097, Codebook Size=8192, Resolution=128x128, Frames=16, Extra Tokens=1002026.01 | 26 | 29.08 | 0.062 | — | — | |
| Omni-DV#Tokens=4096, #Dim.=8, T/P (%)=0.3912026.06 | 27.12 | 28.06 | 0.0637 | 0.9095 | — | |
| AdapTok#Tokens=2048, #Dim.=16, T/P (%)=0.7812026.06 | 27.97 | 26.38 | 0.0599 | 0.8539 | — | |
| STAT-Video#Tokens=997, Codebook Size=8192, Resolution=128x128, Frames=162026.01 | 29 | 28.77 | 0.064 | — | — | |
| AdapTok#Tokens=1024, Codebook Size=8192, Resolution=128x128, Frames=162026.01 | 36 | 25.72 | 0.114 | — | — | |
| AdapTokTokenizer #Params=195M, Generator #Params=633M2026.03 | 36 | — | — | — | 1,024 | |
| OmniTokenizerCompression Spatial (fs)=8×, Compression Temporal (ft)=4×, Input Resolution=256×2562026.06 | 38.15 | 29.2 | — | 0.931 | — | |
| TivTok-T128Video resolution=32×256×256, #Tokens=160, #Dim.=128, Latency (s)=0.202026.06 | 38.49 | 29.05 | 0.0719 | 0.8831 | — | |
| OmniTokenizerTokens (T×H×W)=5 × 16 × 16, Ratio=82026.01 | 42 | — | — | — | — | |
| OmniTokenizerTokenizer #Params=82.2M, Generator #Params=650M2026.03 | 42 | — | — | — | 1,280 | |
| OmniTokenizer-CVToken type=Continuous, Total # tokens=8192, # Frames=172024.11 | 49.5 | 28.3 | 0.081 | 0.913 | — | |
| LARP#Tokens=1024, #Dim.=16, T/P (%)=0.0982026.06 | 51.45 | 25.53 | 0.0973 | 0.8262 | — | |
| CV-VAE#Tokens=4096, #Dim.=4, T/P (%)=0.3912026.06 | 52.43 | 29.47 | 0.0685 | 0.8849 | — | |
| MAGVITTokens (T×H×W)=5 × 16 × 16, Ratio=82026.01 | 58 | — | — | — | — | |
| VideoAR-LTokens (T×H×W)=5 × 8 × 8, Ratio=162026.01 | 61 | — | — | — | — | |
| TivTok-T1024Video resolution=32×256×256, #Tokens=1280, #Dim.=162026.06 | 61.46 | 29.13 | 0.0711 | 0.8857 | — | |
| OmniTokenizer#Tokens=1280, Codebook Size=8192, Resolution=128x128, Frames=162026.01 | 62 | 27.76 | 0.112 | — | — | |
| CV-VAEVideo resolution=32×256×256, #Tokens=8192, #Dim.=4, Latency (s)=1.782026.06 | 64.21 | 29.12 | 0.0692 | 0.8809 | — | |
| PVDM-AEToken type=Continuous, Total # tokens=6144, # Frames=162024.11 | 66.5 | 26.5 | 0.12 | 0.859 | — | |
| OpenSora#Tokens=4096, #Dim.=16, T/P (%)=0.3912026.06 | 67.52 | — | — | — | — | |
| CV-VAEVideo resolution=128×256×256, #Tokens=32768, #Dim.=4, Latency (s)=7.122026.06 | 72.91 | 29 | 0.0729 | 0.8831 | — | |
| CosmosTokenizer-CV*Token type=Continuous, Total # tokens=8192, # Frames=172024.11 | 87.8 | 28.5 | 0.119 | 0.905 | — | |
| TivTok-T128Video resolution=128×256×256, #Tokens=352, #Dim.=128, Latency (s)=0.712026.06 | 92.09 | 26.23 | 0.1057 | 0.821 | — | |
| OmniTokenizer-DVToken type=Discrete, Total # tokens=8192, # Frames=172024.11 | 97.9 | 26.1 | 0.113 | 0.871 | — | |
| CoordTokToken type=Continuous, Total # tokens=1280, # Frames=1282024.11 | 102.9 | 28.6 | 0.066 | 0.892 | — | |
| Cosmos-S#Tokens=512, #Dim.=16, T/P (%)=0.0492026.06 | 104.51 | 28.26 | 0.1046 | 0.8577 | — | |
| Cosmos-Tokenizer-DV#Tokens=1280, Codebook Size=64000, Resolution=128x128, Frames=162026.01 | 140 | — | — | — | — | |
| TATS#Tokens=1024, Codebook Size=16384, Resolution=128x128, Frames=162026.01 | 157 | 24.22 | 0.206 | — | — | |
| TATSTokens (T×H×W)=4 × 16 × 16, Ratio=82026.01 | 162 | — | — | — | — | |
| TATSTokenizer #Params=32M, Generator #Params=321M2026.03 | 162 | — | — | — | 1,024 | |
| LARPToken type=Discrete, Total # tokens=8192, # Frames=162024.11 | 201.3 | 24.3 | 0.142 | 0.806 | — | |
| LARPVideo resolution=32×256×256, #Tokens=2048, #Dim.=16, Latency (s)=1.752026.06 | 226.79 | 23.15 | 0.1757 | 0.7479 | — | |
| ElasticTok#Tokens=1022, Codebook Size=64000, Resolution=128x128, Frames=162026.01 | 230 | — | — | — | — | |
| TATS-AEToken type=Discrete, Total # tokens=8192, # Frames=162024.11 | 249.4 | 23.2 | 0.213 | 0.792 | — | |
| VidTwin#Tokens=1008, #Dim.=4/8, T/P (%)=0.1262026.06 | 388.86 | 28.14 | 0.2414 | 0.8044 | — | |
| ElasticTok#Tokens=1024, #Dim.=16, T/P (%)=0.3912026.06 | 390 | — | — | — | — | |
| OmniTokenizer-CVToken type=Continuous, Total # tokens=1024, # Frames=172024.11 | 396.7 | 23.2 | 0.175 | 0.744 | — | |
| LARPToken type=Discrete, Total # tokens=1024, # Frames=162024.11 | 443.5 | 22 | 0.181 | 0.766 | — | |
| MaskGIT-AEToken type=Discrete, Total # tokens=8192, # Frames=12024.11 | 447.1 | 21.4 | 0.139 | 0.667 | — | |
| OmniTokenizer-DVToken type=Discrete, Total # tokens=1024, # Frames=172024.11 | 509 | 22.2 | 0.201 | 0.703 | — | |
| CosmosTokenizer-CV*Token type=Continuous, Total # tokens=1024, # Frames=172024.11 | 519.6 | 24 | 0.22 | 0.774 | — | |
| CoordTokVideo resolution=128×128×128, #Tokens=1280, #Dim.=82026.06 | 1,108.76 | 27.25 | 0.2346 | 0.7503 | — | |
| PVDM-AEToken type=Continuous, Total # tokens=1152, # Frames=162024.11 | 1,270.1 | 19.1 | 0.333 | 0.563 | — | |
| LARPVideo resolution=128×256×256, #Tokens=8192, #Dim.=16, Latency (s)=22.782026.06 | 3,223.55 | 14.85 | 0.6251 | 0.2924 | — | |
| MAGVIT-AE-LToken type=Discrete, Total # tokens=8192, # Frames=162024.11 | — | 21.8 | 0.113 | 0.69 | — |