- Nov 28, 2020
- 491
- 1,423
Thank you for an update. Some interesting food for thought.Sorry for being late to the party and bringing this up again but your comparison is an oversimplification and the resulting conclusion sadly not sound.
Both JPEG as well as WEBP have image details they struggle with: JPEG suffers on hard borders of uniform color or small lines on uniform backgrounds, WEBP suffers on areas of small details like leaves on a tree or human skin. Given a modern encoder likeYou must be registered to see the linksthe size difference between both formats on equally perceived quality is negligible and depending on image composition JPEG may still draw ahead. So the question still remains and it's answer isn't always WEBP.
That being said, I too second darkhound1's recommendation of using fixed quality WEBP over most JPEG encoders, CrimsonAxisStudio. Though if one tried to optimize for size as well as quality a structural-similarity-based approach using variable quality settings for each image would be the way to go. And while the process may sound rather daunting, it can be automated by tools likeYou must be registered to see the links.
I actually did test this on a relatively new title last autumn which shipped both a PNG- as well as a JPEG-version (the latter with a quality setting of 90, iirc) on its first iterations:
You don't have permission to view the spoiler content. Log in or register now.
I never thoought about it deeply, just used XnConvert (using cjpeg I believe), so with
You must be registered to see the links
it can be different.Still, with large images (like 1920x1080) smoothing over small details (like 1-2px) is not a bad thing, even reducing granularity of the renders. So with high quality settings, IMHO webp seems to be looking typically better for 3D CG. Unless maybe the image is kinda specific, like with many important small details...