I'm on 4.9.11314 (on a H2O), yeah.
I'll read through your thread, and double-check how the top 11px are handled now, compared to how that was done before (a quick glance yesterday told me *something* was still done, but I didn't have time to figure out exactly what).
I'm also using the larger Home/Library covers patches, which I realize is now probably counter-productive with this, since it potentially forces another layer of crappy scaling on top of now-dithered content, which is not that great...
I'll have to ditch those and see if that hunch is correct, too
. (Although, admittedly, I care much more about sleep covers that Home/Library thumbnails).
EDITē: Okay, turns out they're still shown smaller than the resolution of the thumbnails, so, patches are unlikely to change anything, we're always gonna suffer from a downscaling effect there.
EDIT: Okay, quick question for you: for what I gather, you were essentially doing what I'm doing with the "letterbox" option here, except with a dynamic median color instead of black bars, right? Were you also using dithering, or just a full grayscale?
If dithering, was it "automatic" dithering (i.e., palette selection done by the image processing), or a fixed palette?
Were you computing the median before or after the dithering step?
And, if dithering, PNG or JPG?
From what I remember of my experiments doing exactly this on Kindle, waaaaay back in the day, you had to do things in a certain order not to wreck everything (I'd say: remap + dithering -> median -> letterbox, so that the letterboxing uses a color that's on the palette, instead of computing a potential "new" one, which can happen if dithering wasn't done with a fixed palette).
As the state of the ScreenSavers hack attests to today, I personally settled on simple black bars, I was never quite satisfied with the dynamic colors approach. It sometimes looked somewhat better, but it sometimes looked alien. Black bars have the advantage of never being really terrible
. It was certainly interesting to experiment with, though
.