TBH I can't see any artifacts, but I assume if the jpg is 2MB and the png is 20MB, the lossy compression had to make some cuts somewhere. (After checking the standard ratio seems to be 10:1 for lossy compression).
TBH I can't see any artifacts, but I assume if the jpg is 2MB and the png is 20MB, the lossy compression had to make some cuts somewhere. (After checking the standard ratio seems to be 10:1 for lossy compression).
The easiest places to see jpeg artifacts here IMO are Frieren's earring and the highlights on Stark's coat. JPEG is especially bad with reds. Zooming in you can see minor color munging and blurring elsewhere. It does appear this is a legitimate better quality version and not lossy-lossless, at least compared to the twitter child.
It's a noisy image (as in lots of effectively random color differences between nearby pixels, looking like a grainy texture) which is basically the hardest thing to losslessly compress, so it makes sense the lossless is larger than average for the resolution because it has to save all those individual colors. A lossy algorithm can throw away some of those differences. PNG does attempt to compress (losslessly, like a zip file). PNGs with flat colors can actually be quite small. As a gross oversimplification, a white square can be "250x250 white" instead of "white white white white white white... (repeat 62500 times)". Practical example: post #7437245 has more pixels than this one (167%) but is less than 1/4 the filesize.
I don't think there's a reliable "standard ratio" for lossy compression. It will depend on what's in the image, and the algorithm and compression level -- jpeg has a quality level (the twitter child here is 85) and even lossy encoders usually have levels, which don't affect quality but affect the time it takes to compress (and sometimes time it takes to decompress too). That's not to say you can't make statistical generalizations, if you're trying to estimate required storage space or something, but just that filesize is not a reliable indicator of quality level or how lossy an image is. Even between two jpegs, filesize usually shows you which is higher quality, assuming the same pixel size and the same source image, but even that rule can be broken if you make a jpeg out of another jpeg, so that the artifacts from the first time are compressed too, resulting in a larger file with lower quality at the same quality level setting.
I nearly cried when I saw this happen in Ordinal Scale, having Asuna do Mother's Rosario and have Yuuki appear was a great cameo and homage to her. T-T
I nearly cried when I saw this happen in Ordinal Scale, having Asuna do Mother's Rosario and have Yuuki appear was a great cameo and homage to her. T-T
Me too. I shed tears at the end of the Mother's Rosario arc.
I shed tears at the end of Mother's Rosario arc. I have never ever in my life shed tears for fictional characters except for 4 Yuuki, Eugeo and for Toradora Aisaka Taiga and Takasu Ryuuji romance.
