Jon Sneyers

This seems to be a larger difference than expected. Can you share the original jpeg for this example? I want to try to reproduce this.

That does not do lossless jpeg recompression at all. That decodes the JPEG to pixels, downscales the image 2x, then encodes the result losslessly. Obviously downscaling an image 2x is...

Question still is which one is closer to the actual jpeg spec, i.e. using mathematically exact, infinite precision arithmetic. As far as I understand, the existing jpeg decoders like libjpeg-turbo...

Mozjpeg/libjpeg-turbo float is doing arithmetic during iDCT with floats, but it's still converting int16 dct coeffs to uint8 ycbcr values. I don't think any part of libjpeg-turbo uses float buffers...

If you do --strip, then it will drop the jpeg bitstream reconstruction so djxl will decode to rgb pixels and then do a reencode as q95 jpeg. Without the --strip,...

+1 Encode memory usage should be improved, at least for the lower effort settings.

JXL is always tile-based, but it's a bit tricky since some things like the filtering does go across tile boundaries. The main thing is that the current encoder tries to...

Looks like there's an issue with the byte order of FLOAT16 on big-endian architectures... Can you check if the problem still happens with the current git version?

JPEG XL splines are specified/rendered the way they are, if we would change that, it would change how different versions of the decoder would decode an image. They are 'blurry'...

For impure colors, it doesn't work since you'll get something like this: https://jxl-art.surma.technology/?zcode=C89MKclQMDI14PJIzUzPKAEzuYILcjLzUgNLE_NKMqsSSzLz8xxTskqLS3JT80oUDKDSXAoKBpRBoAm6BnpGlBtiSLERhpR7xdQAiIC0kQGY4ZqXAg0n6gaXGeWGGFNshBk1QsvQABZcIBZtwosaiYtyZ5hQJbjgqcuQZsnLlAohTrkZxtQILiN46jJCSV1cugrBqSUKhkYWAA (i.e. the pure version of the color creeps in, and only the edges are actually in the...