Dithering is the trick of adding a tiny amount of controlled noise to disguise the limitations of a low bit depth. The eye is bothered by visible bands of stair-stepped colour on smooth gradients (banding), but it's surprisingly tolerant of fine random noise — even when the noise contains less information than the bands it replaces. By adding the right kind of noise, dithering breaks up the visible bands into a much more pleasant texture that looks like a slightly grainy gradient instead of obvious steps.
The technique has been around since the 1970s for printing greyscale images on black-and-white printers, and was a workhorse of digital audio mastering in the CD era. In modern video, dithering shows up in two main places. First, at the encoder — modern video encoders (x264, x265, SVT-AV1) can apply dithering during quantization, which trades a tiny amount of extra bitrate for noticeably less banding on dark gradients. Second, at the decoder/display — when 10-bit content has to be shown on an 8-bit screen, the player can dither the conversion instead of just clipping to the nearest 8-bit value, dramatically improving the perceived quality of HDR content on lesser displays.
For product teams, dithering rarely needs to be tuned manually but is worth knowing about. Two practical signs it matters: if your dark scenes look like coloured topographic maps with visible elevation lines, ask the encoding team if dithering is enabled. If your 10-bit HDR content looks "posterised" on some viewers' devices, the SDR-conversion path in the player is probably dithering badly or not at all. The fixes are cheap; the visible quality wins are large.
