Distortion is the technical term for "how much the compressed video differs from the original". Every lossy codec — H.264, HEVC, AV1, VVC — introduces some distortion when it discards information to make files smaller; the encoder's job is to keep that distortion below what the human eye can notice. The term shows up constantly in encoder documentation, codec research and metric formulas, and it's the "D" in the famous rate-distortion trade-off equation (J = D + λ·R) that drives every encoding decision.
Distortion can be measured many ways, and which one you choose changes the conclusions. Pixel-level distortion (MSE, PSNR) measures raw numerical difference between source and compressed pixels — easy to compute, weakly correlated with how humans perceive quality. Structural distortion (SSIM, MS-SSIM) measures how much the structure and texture of the picture changed — better correlated with perception. Perceptual distortion (VMAF) is trained specifically to predict what humans rate as distortion in subjective tests. The same encode can show low MSE distortion but high VMAF distortion (or vice versa) depending on what kind of artefacts it introduced.
For a product team, distortion is the vocabulary engineering and research papers use for "quality loss" — useful to know when reading codec papers or vendor benchmarks. The practical takeaway: don't trust a single distortion metric. PSNR alone can mislead because it doesn't match human perception; VMAF alone can be gamed by encoders that target it specifically; subjective testing (ACR/MOS) is the ground truth but expensive. Serious quality analysis reports several distortion metrics side by side and looks at where they agree and disagree.
