Video codecs compress footage using two main frame types:
I-Frames (Intra / Keyframes)
Complete image snapshots. They contain all pixel data and serve as reference points.
P-Frames (Predicted / Delta)
Only store the differences from the previous frame — motion vectors and small corrections.
Datamoshing exploits this by removing I-frames, so P-frames apply their motion data to the wrong reference image. The result: pixels smear, warp, and melt along motion paths, creating hallucinatory visual artifacts.
This app uses the WebCodecs API with the VP8 codec to encode → intercept → manipulate → decode frames in real-time right in your browser.