See https://github.com/libjxl/libjxl/pull/466 https://github.com/libjxl/libjxl/pull/466 for more details. As far as I understand it, the synthetic mode compres

See https://github.com/libjxl/libjxl/pull/466 https://github.com/libjxl/libjxl/pull/466 for more details. As far as I understand it, the synthetic mode compresses better for (as you can guess) synthetic content. So if you're encoding a mixed content screenshot, you'll get better quality and size by encoding the photos and UI separately.