AI Paper Atlas

Research map / Video / Video generation

Flash-VAED: Plug-and-Play VAE Decoders for Efficient Video Generation

Lunjie Zhu, Yushi Huang, Xingtong Ge, Yufei Xue, Zhening Liu, Yumeng Zhang, Zehong Lin, Jun Zhang

ICML 2026 · Video › Video generation · approach: Efficient video diffusion models · 0 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 independence-aware channel pruning instead of random or L1-norm pruning (ablation)Flash-VAED · PSNR over training stepsPSNRImproved
add pre-pruning retained channel enhancement (expressivity loss)Flash-VAED channel pruning · feature reconstructionR2: 0.87-0.93 -> exceeding 0.98Improved
replace CausalConv3D with 2D convolutions in higher-resolution layers (ablation)Wan 2.1 VAE decoder · UCF-101 reconstructionlatency and PSNR: latency 396 -> 307 ms; PSNR 32.42 -> 32.99 (up0 vs up3)Improved
reduce prune ratio from 1/2 to 1/4 (ablation)Flash-VAED-LTX · UCF-101 reconstructionFPS / PSNR: 625.00 -> 1167.99 FPS; PSNR 32.5421 -> 32.2400Mixed
reduce prune ratio from 1/4 to 1/8 (ablation)Flash-VAED-LTX · UCF-101 reconstructionFPS / PSNR: 1167.99 -> 1907.23 FPS; PSNR 32.2400 -> 31.0786Mixed
initialize projection layer with derived W instead of random (ablation)Flash-VAED-LTX phase 3 · UCF-101 reconstructionPSNR: 30.3061 -> 32.2400Improved
add three-phase feature distillation (vs none) (ablation)Flash-VAED-LTX · UCF-101 reconstructionPSNR: 30.7957 -> 32.3768Improved
distill only deep layers instead of full three-phase (ablation)Flash-VAED-LTX · UCF-101 reconstructionPSNR: 30.7957 -> 30.8065 vs no distillationNo clear effect
distill only shallow layers instead of full three-phase (ablation)Flash-VAED-LTX · UCF-101 reconstructionPSNR: 32.3768 -> 31.2386Hurt
swap original VAE decoder for Flash-VAED in Self Forcing-Wan 1.3BSelf Forcing-Wan 1.3B · end-to-end generation, VBench-2.0latency: 10.14s -> 7.41s (27%)Improved
swap original VAE decoder for Flash-VAED in FastVideo-Wan 1.3BFastVideo-Wan 1.3B · end-to-end generation, VBench-2.0latency: 9.66s -> 6.15s (36%)Improved
use LightVAE decoder instead of Flash-VAED in generation pipelinesFastVideo-Wan 1.3B · VBench-2.0VBench-2.0 qualityHurt

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