VAE (latent autoencoder)
Compresses images into a small latent space where diffusion actually happens, then decodes the result back to pixels.
Introduced inDec 2013Auto-Encoding Variational Bayes
Latent diffusion models never diffuse pixels directly. A variational autoencoder first compresses the image ~8× per side into a compact latent tensor, the diffusion process denoises in that space, and the VAE decoder renders the final latent back into pixels. This is the trick that made high-resolution diffusion affordable, and it is why generation graphs end with a VAE decode stage.
Adoption over time
Share of new models that have included a VAE over time.
See it in real models
Open any of these on hfviewer to find this block in the interactive architecture graph.
Comfy-Org/MiniMax-H3↓ 11.8M · ♡ 1kOpen in visualizer
MiniMaxAI/MiniMax-H3image-text-to-video · ↓ 2.0M · ♡ 4kOpen in visualizer
Lightricks/LTX-2.3image-to-video · ↓ 1.7M · ♡ 2kOpen in visualizer
stable-diffusion-v1-5/stable-diffusion-v1-5text-to-image · ↓ 1.5M · ♡ 1kOpen in visualizer
stabilityai/stable-diffusion-xl-base-1.0text-to-image · ↓ 1.5M · ♡ 8kOpen in visualizer
stabilityai/sdxl-turbotext-to-image · ↓ 1.3M · ♡ 3kOpen in visualizer
microsoft/TRELLIS.2-4Bimage-to-3d · ↓ 1.3M · ♡ 1kOpen in visualizer
black-forest-labs/FLUX.2-devimage-to-image · ↓ 1.1M · ♡ 2kOpen in visualizer
Tongyi-MAI/Z-Image-Turbotext-to-image · ↓ 934k · ♡ 5kOpen in visualizer
microsoft/VibeVoice-ASRautomatic-speech-recognition · ↓ 682k · ♡ 1kOpen in visualizer
microsoft/VibeVoice-Realtime-0.5Btext-to-speech · ↓ 660k · ♡ 1kOpen in visualizer
Lightricks/LTX-Videoimage-to-video · ↓ 560k · ♡ 2kOpen in visualizer
Browse all 216 models with this in the catalog →
Related concepts
hfviewer renders the full architecture of 3,500+ Hugging Face models as interactive graphs. Hover any block to see what it does, with this model’s real numbers.
Browse all model graphs →