@fastidious On the loss of quality on image processing / resizing… I think it might have something to do with the resampling that occurs during the resize operation. On the backend
yarnd we use the gift library for most processing (mostly just
ResizetoFit()). PNG itself is a lossless image format with selectable compression levels depending on output size and decode speed, so I think the culprit is the resampling…
I’m not an expert though in resampling algorithms when it comes to resizing images. So I’m not really sure whether the Resampling we use
gift.LanczosResampling is the best choice here 🤔
@prologic I will run a small research to see what’s would the best resampling be. I will be looking something of a compromise, whilst trying to achieve better quality.
@fastidious Sounda good 👌
@prologic so I checked. It seems we are already using the best resampling method. So… hmm, yeah. Other than resizing the images less—say, instead of 850px, resize to 1000px, if bigger than 1000px—I don’t see any other way to get better results.
@fastidious To be honest I’m not sure how even using an “original size” image and doing CSS size to fit or even on any app really doesn’t also suffer from loss of quality due to resampling. Resampling is a pretty common algorithm to apply to any image resizing I thought? 🤔
@prologic this is easy to test. I will pull one original, host it, and link it on a twt. Then we will compare it to the uploaded version. Granted, it will be a visual comparison only.
@fastidious Ok 👌