AppleTV has their own graphic design for now-playing data, so it shows some stuff, but not as nicely.
On Chromecast, we control the rendering, since a Chromecast is (more or less) an embedded web browser, so we did our own design for that screen in @RBM’s screenshot.
We have some work to do on this, too. Sourcing or crowd-sourcing better artwork will help, but the other thing we really need content-aware cropping/positioning of images on the display side.
We call this the “foreheads problem”:
The idea is to identify the most salient portion of the image (red bounding box), the center-of-mass of salience (red dot), and any human faces (blue boxes) in the image.
Using this extracted information, we can come up with cropping rules that avoid chopping faces in half and focus attention on the most interesting parts of an image.
The exact crop rectangles different from situation to situation, but the idea is that for each display scenario–whether it’s a 16:9-ish like Chromecast, or a square image in a grid, or a 4:3 on the artist browser, we’ll be able to intelligently crop the image for the situation.
Anyways–we have the computer vision/feature extraction part working well, but there is a lot of “plumbing” work involved in getting it deployed for real and then distributing the benefits to all the relevant places in the product.