YouTube Banner Safe Area Guide — one upload, several crops
A YouTube channel banner is the most extreme multi-crop case of any profile image on any platform. You upload one wide picture, and YouTube then shows a different slice of it on every surface it appears on. The living-room television app draws the largest slice, closest to the whole uploaded frame. Desktop web takes a shorter, wider band out of the middle. A tablet takes a narrower band again. A phone shows the smallest slice of all — a modest central rectangle. Each of those is a nested crop of the same frame, which is why a banner that looks balanced in an editor so often arrives on a phone with the logo gone: the editor showed you the whole rectangle, and no surface ever displays the whole rectangle.
The safe area is an intersection, not a box you can look up
Because the surfaces crop differently, there is no single YouTube banner safe area to memorise off a spec sheet. The region that appears intact everywhere is the geometric intersection of every crop — the part inside all of them at once — and it is always smaller than any individual crop. Because the YouTube crops are nested, the phone slice will usually be that intersection, but this tool does not assume it: it computes the overlap from whatever crops are switched on, so the answer stays right if the nesting ever stops holding and so that isolating two surfaces still produces the correct box.
Toggling the device crops is the point of the tool rather than a convenience. Switch everything off except the television and watch the safe box balloon; switch the phone back on and watch it collapse. That collapse is the whole lesson, and it is the reason a corner logo and a bottom-edge upload schedule so rarely survive. The box is reported both as x, y, width, height in pixels at the frame’s reference resolution — numbers you can type straight into an image editor — and as percentages, for tools that work proportionally.
Cropped away and covered over are different failures
Some of the banner is not cropped at all, but obscured. The channel name, the avatar, the subscribe controls and other interface chrome are composited over the banner on the channel page, and some surfaces darken a portion of it. Those pixels are rendered and then hidden, which is a different problem from pixels that were never rendered, so the overlay draws them differently: crops are dashed outlines that shrink the safe box, while obstructions are hatched shapes drawn over the top that never shrink it. An obstruction does not stop the rest of that surface showing the banner — it just means nothing readable can sit underneath it.
Fractions, not pixels
Every region is stored as a fraction of the uploaded frame rather than as an absolute pixel box. A banner is rendered at whatever width the viewport happens to be, so a rectangle measured in pixels on one display is wrong on every other one. Keeping the geometry resolution-independent also means the overlay is correct at any canvas size, and that YouTube restating its nominal upload size does not silently invalidate the drawing. The reference pixel dimensions are carried once per frame, for the readout alone.
Your artwork stays on your machine
Loading a banner is optional — the geometry is drawn on a placeholder frame with no image at all. If you do load one, it is decoded, drawn and re-encoded entirely by your own browser; there is no server to send it to and no field for a remote address, because a cross-origin image would taint the canvas and break the download. You can export the annotated guide, or the plain banner cut to a single device crop, which shows exactly what that surface displays.