Back to wiki Architecture

Wan-VAE

Wan-VAE is the 3D causal compressor in front of the DiT. Wan2.1 used 4×8×8. Wan2.2-VAE uses 4×16×16 (information compression rate 64) and is what makes TI2V-5B a realistic 720P@24 fps model on a 4090.

4×16×16TI2V-5BPSNR
On this page 5
Official VAE comparison table including Wan2.1-VAE and Wan2.2-VAE
Official reconstruction table. Wan2.2-VAE: 4×16×16, feature dim 48, compression rate 64, PSNR 33.223, SSIM 0.922, LPIPS 0.022.

Read the official row

VAET×H×WFeature dimInfo ratePSNR↑SSIM↑LPIPS↓
SVD1×8×844828.3070.8160.067
Cosmos4×8×8164829.6320.8630.132
Hunyuan4×8×8164833.2980.9220.023
CogVideoX1.54×8×8164832.9550.9220.040
Wan2.1-VAE4×8×8164832.2220.9110.026
Wan2.2-VAE4×16×16486433.2230.9220.022

The 2.2 VAE is more compressed spatially (16 vs 8) with a wider latent channel (48 vs 16). Reconstruction still matches or beats the 4×8×8 pack on SSIM and LPIPS in this table. That is the claim to repeat — not a homemade “64× better” slogan.

Patchify on TI2V-5B

README: with an extra patchification layer, TI2V-5B’s total compression reaches 4×32×32. Combined with a 5B dense DiT, a 5-second 720P clip is quoted as under 9 minutes on a single consumer GPU without extra compiler tricks.

A14B still uses the earlier VAE family

The high-compression VAE is called out for the 5B hybrid. The 27B MoE T2V/I2V models keep a compatible latent with the 14B-scale stack (Wan2.1-style 4×8×8). Mixing VAE weights across those families will decode garbage.

FAQ

Why 4 in time?
Causal 3D VAE: four frames (plus a boundary) collapse to one temporal latent. That is why video lengths are 4n+1 frames.
Can I decode 2.2 latents with a 2.1 VAE?
No. Channel count and spatial stride differ (48 vs 16, 16 vs 8).

Primary sources

Checked against Wan-Video/Wan2.2 README · Aug 28, 2026

© 2026 wan2.video