Put a plain red title card in a video and something ugly happens: the letters grow fuzzy halos, the edges bleed like wet ink, and thin strokes dissolve entirely — while the same text in white or gray stays crisp at the same size. Video creators learn it as folklore: never use pure red text on video. But the rule is not superstition. It is the visible fingerprint of a deliberate engineering compromise baked into nearly every video you have ever watched: color is stored at much lower resolution than brightness, and red text is where that discount shows. It has stubborn cousins on the display side, too — the way white text on a black background can shimmer and glow — but this one happens in the video file itself, before your screen ever gets involved.
The system has a name, chroma subsampling, and it exists because of a fact about your eyes. Human vision resolves fine detail mostly through brightness, not color. You can read a black-and-white photograph down to tiny features, but if you strip all brightness information away and leave only color, the scene becomes mush. Video engineers noticed this decades ago and built it into the foundations of digital television.
Two channels, one budget
Digital video splits every frame into a luminance channel — the black-and-white detail — and two chrominance channels carrying the color differences. The labels you may have seen, 4:4:4, 4:2:2 and 4:2:0, describe how many pixels share each channel's samples. In 4:4:4, every pixel keeps its own color. In 4:2:2, color is sampled every two pixels horizontally. In 4:2:0 — the format of broadcast digital video, streaming, and the default of essentially every consumer camera and codec from DVDs to phone video — color is stored at half resolution horizontally and half vertically, one color sample doing the work of four pixels. The brightness channel, meanwhile, keeps every pixel.
The family tree is old and official. The international standards that first defined digital studio television in the 1980s wrote subsampled chroma into broadcast video itself, and modern compression standards such as H.264, which carries the bulk of today's streamed video, specify 4:2:0 as the working format for its mainstream profiles. This is not a corner case; it is the water video swims in.
The bet mostly pays off. Because your eyes have far fewer neurons devoted to spatial detail in color than in brightness, a frame can lose three-quarters of its color resolution and still look like a photograph of reality — skies stay blue, faces stay warm. Research on subsampling continues precisely because the loss is not always invisible: studies of HDR and high-resolution video find viewers can prefer systems that restore or refine chroma, especially on large bright displays, and grading suites work in higher chroma resolutions for exactly this reason.
Why red suffers first
Subsampling punishes whatever lives only in the color channels. Consider the two title cards. White text on black is a pure brightness event: the luminance channel swings hard at every letter edge, and since luminance keeps full resolution, the edge survives at pixel sharpness. Red on white is the opposite: the two colors are nearly equal in brightness, so luminance sees almost nothing at the boundary. All the information that says "letter ends here" lives in chroma — and chroma is running at quarter resolution. The letter's outline is, quite literally, drawn with a brush one-quarter the width of the canvas.
On playback, the display stretches each color sample across four pixels to fill the frame, and the red boundary becomes a smear two to four pixels wide. Screens add their own layer: a screen renders every color from three subpixels the way a water droplet reveals directly, and at small text sizes the subsampled smear interacts with the subpixel grid, sometimes leaving a fringe on one side of a stroke and a gap on the other. Magnified, red video text looks less like typography than like a weather map of red probability.
Working with the discount instead of against it
Once you know the mechanism, the folklore writes itself into rules with reasons. Larger font sizes give the quarter-resolution brush more room to trace a recognizable shape. Slightly darker or lighter reds — a crimson toward black, a coral toward white — reintroduce a luminance difference at the edge, handing part of the outline back to the sharp channel. Outlines and drop shadows work for the same reason: they are brightness events. And text rendered in white, off-white, or yellow on dark backgrounds sidesteps the problem entirely, which is why subtitles have looked the way they look for forty years.
The smear also explains a puzzle every screen photographer has met: photographing a video frame can look worse than playing it, since screenshots mishandle the brightness arithmetic of modern video in their own way — capture pipelines and playback pipelines each take their cut.
What the blur quietly demonstrates is something vision scientists have known since the nineteenth century: the world you see is not uniformly detailed. Your own eyes subsample color, coarsely and early, and the video system merely copies your specifications back at you. When two nearly-equal-in-brightness colors sit side by side, you are seeing with the low-resolution channel — and if you want to know how sharp your full-color judgment really is at small differences, a run through Spot the Different will tell you quickly. The blur was always in your eye first; the engineers just made it official.
every pixel keeps its own color
two pixels share one color
four pixels share one — the video default
Edges that exist only in color, like red text on white, blur by exactly the amount the stored color resolution allows.