https://en.wikipedia.org/wiki/Image:Col … 24x768.gif

You can't link a webpage as a picture.  :)  For some reason, linking to the image directly produces a low-quality version in my browser but here you go anyway:

Reviving this old wallpaper thread because the problem vilem ran into in 2007 still bites people making Colemak desktop images today: the picture gets scaled to fit the window (or the screen), and a layout diagram with small key legends turns to mush very quickly.

A few things that make a Colemak wallpaper hold up on modern screens:

1. Start from vectors, export per resolution. The original was made in Illustrator, which is exactly right. Keep the master as SVG/AI and export a separate PNG for each common size (1920x1080, 2560x1440, 3840x2160, plus 16:10 variants) instead of letting the OS stretch one GIF. GIF in particular is limited to 256 colours, which is why the Wikipedia version looked rough once resized.

2. Put the keyboard in the lower-middle third. Desktop icons usually sit on the left, the taskbar/dock at the bottom or side. A layout placed slightly below centre with generous margins stays readable and does not get covered.

3. Use high-contrast legends and colour-code the moved keys. The most useful wallpapers I have seen on this forum highlight the keys that differ from QWERTY (E, R, T, Y, U, I, O, P, S, D, F, G, J, K, L, N and the semicolon), or group them by finger. Muted background, saturated key colours, and legends at least ~2% of screen height.

4. Make a learning version and a "finished" version. While learning, a big full-layout image helps. Once you are past the Tarmak steps, many people prefer a subtle wallpaper where the layout is faint and only the home row stands out, so it does not become a crutch.

5. Backgrounds. If you want the layout over a photo rather than a flat colour, pick a photo with a calm area where the keyboard goes, darken or blur that area, and overlay the vector layout on top. If you want to restyle a photo first (turn it into a watercolour or a minimal flat-colour scene, or extend a narrow photo to 21:9 so it fills an ultrawide monitor), an image to image ai editor can do that in the browser from the photo plus a short prompt while keeping the composition, and then you add the crisp vector layout on top in Inkscape/Illustrator. Never let an image tool redraw the key legends themselves; text is where generated images still go wrong, so the legends should always come from the vector file.

If anyone still has the original 1024x768 files from this thread, it would be nice to see them re-exported at modern sizes and collected in one place.