List your company
ResearchFormerly WizWand

Video Generation on UCF-101

58FVDMAGVITv2
-16.36485.57987.51,489.43Dec 10, 2024Mar 1, 2025May 21, 2025Aug 10, 2025Oct 30, 2025Jan 19, 2026Apr 10, 2026
Updated 4mo ago

Evaluation results

MethodLinks
MAGVITv2Type=Mask., Params=307M2024.1258————
ELT (6N x 6L)Class=class-conditional, # params=76M, # steps=24, GFlops=∼13k, Pretrained on additional large video data=true2026.0460.8———87.88
ELT (6N x 4L)Class=class-conditional, # params=76M, # steps=12, GFlops=∼4.3k, Pretrained on additional large video data=true2026.0472.8———88.27
MAGVITType=Mask., Params=306M2024.1276————
MAGVIT-LClass=class-conditional, # params=306M, # steps=12, GFlops=∼4.3k, Pretrained on additional large video data=true2026.0476———89.27
Make-A-VideoClass=class-conditional, # params=≫3.5B, # steps=≫250, Pretrained on additional large video data=true2026.0481———82.55
ACDIT-H-LTType=AR+Diff., Params=954M, Training Strategy=Longer Training2024.1290————
MaGNeTSClass=class-conditional, # params=306M, # steps=12, GFlops=∼1.7k, Pretrained on additional large video data=true2026.0496.4———88.53
PAR-4xClass=class-conditional, # params=792M, # steps=323, Pretrained on additional large video data=true2026.0499.5————
PAR-16xClass=class-conditional, # params=792M, # steps=95, Pretrained on additional large video data=true2026.04103.4————
ACDIT-HType=AR+Diff., Params=954M2024.12104————
MAGVITv2-ARType=AR, Params=307M2024.12109————
ACDIT-XLType=AR+Diff., Params=677M2024.12111————
SVDResolution=1024 × 576, NFE=50, TFLOPs=45.43, Latency (GPU)=376, Latency (Phone)=OOM2024.12149————
MobileVDResolution=512 × 256, NFE=1, TFLOPs=4.34, Latency (GPU)=45, Latency (Phone)=17802024.12171————
VideoFusionType=Diff., Params=510M2024.12173————
SF-VResolution=1024 × 576, NFE=1, TFLOPs=45.43, Latency (GPU)=376, Latency (Phone)=OOM2024.12181————
MobileVD-HDResolution=1024 × 576, NFE=1, TFLOPs=23.63, Latency (GPU)=227, Latency (Phone)=OOM2024.12184————
OmniTokenizerType=AR, Params=650M2024.12191————
MattenType=Diff., Params=853M2024.12211————
MAGVIT-ARType=AR, Params=306M2024.12265————
Video-LaVITType=Diff., Params=7B2024.12281————
AnimateLCMResolution=1024 × 576, NFE=8, TFLOPs=45.43, Latency (GPU)=376, Latency (Phone)=OOM2024.12281————
MMVGType=Mask., Params=230M2024.12328————
TATSType=AR, Params=331M2024.12332————
TATSClass=class-conditional, # params=321M, # steps=1024, Pretrained on additional large video data=false2026.04332———79.28
MagDiffType=Diff., Params=2B2024.12340————
Make-A-VideoClass=class-conditional, Pretrained on additional large video data=false2026.04367———33
LVDMType=Diff., Params=437M2024.12372————
CCVS+StyleGANClass=unconditional, Pretrained on additional large video data=false2026.04386———24.47
TATSClass=unconditional, # params=321M, # steps=1024, Pretrained on additional large video data=false2026.04420———57.63
SVDResolution=512 × 256, NFE=50, TFLOPs=8.60, Latency (GPU)=82, Latency (Phone)=OOM2024.12476————
LatteType=Diff., Params=674M2024.12478————
DIGANClass=unconditional, # steps=1, GFlops=∼148, Pretrained on additional large video data=false2026.04577———32.7
CogVideoType=AR, Params=9.4B2024.12626————
CogVideoClass=class-conditional, # params=9.4B, Pretrained on additional large video data=true2026.04626———50.46
LADDResolution=1024 × 576, NFE=1, TFLOPs=45.43, Latency (GPU)=376, Latency (Phone)=OOM2024.121,894————
UFOGenResolution=1024 × 576, NFE=1, TFLOPs=45.43, Latency (GPU)=376, Latency (Phone)=OOM2024.121,917————
AnimateDiff (V3)zero-shot=true2023.11—46.224.189.8—
DVD-GANClass=class-conditional, # steps=1, Pretrained on additional large video data=false2026.04————32.97
I2VGen-XLzero-shot=true, Input Type=text&image-to-video, Training Videos=10M2023.11—44.121.389.4—
MagDiffzero-shot=true, Input Type=text&image-to-video, Training Videos=5.3M + 76K2023.11—50.825.490.2—
RaMViDClass=unconditional, # params=308M, # steps=500, Pretrained on additional large video data=false2026.04————21.71
StyleGAN-VClass=unconditional, # steps=1, Pretrained on additional large video data=false2026.04————23.94
Video DiffusionClass=unconditional, # params=1.1B, # steps=256, Pretrained on additional large video data=false2026.04————57