AI Paper Atlas

Research map / Video / Video generation

Progressive Growing of Video Tokenizers for Temporally Compact Latent Spaces

Aniruddha Mahapatra, Long Mai, David Bourgin, Yitian Zhang, Feng Liu

ICCV 2025 · Video › Video generation · approach: Efficient video diffusion models · 3 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use progressive growing instead of direct full-model training for 8x temporal compressionMagViT-v2 tokenizer · MCL-JCV 8x8x8 z_dim 8PSNR / rFVD: 28.47 -> 30.26; 219.36 -> 96.85Improved
use progressive growing instead of direct training for 16x temporal compressionMagViT-v2 tokenizer · MCL-JCV 8x8x16 z_dim 8PSNR / rFVD: 26.15 -> 28.31; 255.51 -> 183.4Improved
use progressive growing for 8x compression at z_dim 16MagViT-v2 tokenizer · MCL-JCV 8x8x8 z_dim 16PSNR / rFVD: 28.51 -> 32.06; 135.57 -> 63.64Improved
use progressive growing for 16x compression at z_dim 16MagViT-v2 tokenizer · MCL-JCV 8x8x16 z_dim 16PSNR / rFVD: 28.51 -> 30.02; 135.57 -> 115.25Improved
add residual embedding with AdaNorm and subsampled key-frame encoding on top of naive progressive training (ablation)ProMAG 8x · MCL-JCV 512x512PSNR / rFVD: 30.35 -> 32.06; 88.18 -> 63.64Improved
naive progressive training (add bottleneck blocks to pretrained 4x model) vs training from scratch (ablation)MagViT-v2 8x · MCL-JCV 512x512PSNR / rFVD: 28.51 -> 30.35; 135.57 -> 88.18Improved
add GAN loss during progressive growing (8x) (ablation)ProMAG · MCL-JCV 8x8x8PSNR / rFVD: 32.06 -> 31.78; 63.64 -> 53.18Mixed
add GAN loss during progressive growing (16x, z_dim 16) (ablation)ProMAG · MCL-JCV 8x8x16PSNR / rFVD: 30.02 -> 29.67; 115.25 -> 99.64Mixed
use 8x model on full frames instead of 4x model on subsampled frames plus 2X interpolation (ablation)ProMAG · MCL-JCV 512x512PSNR / rFVD: 30.68 -> 32.06; 230.16 -> 53.77Improved
use 16x model instead of 4x model on subsampled frames plus 4X interpolation (ablation)ProMAG · MCL-JCV 512x512PSNR / rFVD: 28.42 -> 29.6; 341.14 -> 147.48Improved
progressive growing instead of full-model training: training time per iteration (8x)ProMAG · training efficiencytime per iteration: 0.92s -> 0.42s; 4.25 days -> 1.94 daysImproved
progressive growing instead of full-model training: training time (16x) (ablation)ProMAG · training efficiencytraining time: 6.01 days -> 2.21 daysImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy