Image models don't paint pixel by pixel — a VAE crushes the picture into a tiny block of numbers, then decodes it back. Shrink the latent grid and watch detail dissolve; sample from noise and watch it hallucinate.
Live preview · launch for the interactive version
Drag the latent size down and watch a clean image collapse into a blocky code.
The decoder rebuilds a full image from the shorthand — fidelity bounded by the code's size.
Fill the latent grid with random noise and the decoder hallucinates something new from nothing.
Predict which preset survives compression best: gradients, hard edges, or fine detail.
Hold the preset; change only the compression ratio.
Which features blur first? Which hold longest?
Say what compression bought and what it cost.