Codec-data manipulation creating motion artifacts and color streaks — exploits compression algorithms intentionally. Signature glitch aesthetic for music videos and experimental cinema.
You're working with compressed video data and suddenly notice: motion smears across multiple frames, colors blur into blocks, individual image elements duplicate or distort geometrically. This isn't a camera error – this is datamoshing, and you can control it deliberately.
Datamoshing occurs when you manipulate the internal reference frames of a compressed video. Codec formats like H.264 or ProRes don't store every frame completely – they store so-called I-frames (key frames) and derive P-frames and B-frames from them, which only contain motion vectors and differences. If you disrupt this structure – for example, by combining P-frames from different sequences or deleting I-frames – visible artifacts emerge: motion data is applied to entirely wrong spatial areas, color information smears across object boundaries.
This doesn't happen on set, of course. You create datamoshing intentionally in post – either through specialized tools like Avidemux or by directly editing video files at the raw data level. Some work with custom scripts in Python or similar. The effect is particularly pronounced with high motion or fast cuts, as the codec errors are then most visible.
In practice, you need two things: first, uncompressed or minimally compressed source material – the higher the bitrate, the more controllable the result. Second: patience. Every change to the codec data requires re-encoding, and not every manipulation leads to the desired look – much of it is trial and error.
The aesthetic itself: motion trails that extend beyond image borders, blocky color zones, duplications of subjects. It looks deliberately digital, glitchy, almost like corrupted or aged files. Some directors use it for dream sequences or psychological disorientation. Others employ it as a visual leitmotif in music videos or experimental works. The effect has its own presence – it's not motion blur, not tracking errors, but something genuinely different.
Important: Datamoshing works better with older, less sophisticated codecs. Modern 10-bit material can lead to uncontrollable results. And the look is no longer a trend – anyone using it should have a narrative justification, not just nostalgia for YouTube artifacts from 2010.