I'm working with UHD 100Mb/s 4:2:0 HEVC files. They are a pain to edit with, so I wanted to transcode them into a mezzanine codec. I'm using Windows, so that'd be DNxHD as far as I know.
The HQ flavor of DHxHD is 7x the size of the original file, which is hella bloat. The LB version is only 1.4x the size. Would that be lossless since the bucket's not overflowing (so to speak), or are there other factors that could degrade my footage?
None of these codecs are "lossless" Eugene. Some may be visually lossless, though. LB, however, is akin to ProRes Proxy. It's not meant for delivery but SQ is. You may want to check out a test with that format. Let us know what you end up using.
I did do a head to head for LB SQ and HQ and LB definitely degrades the quality noticeably. SQ did not, but it's still 4.5x the size.
I had two questions from all this, if anyone can answer them:
1) Is there any meaningful difference from visually lossless codecs and lossless codecs? I've never seen instances of how a visually-lossless codec reduces the quality of the final image, so being able to see that would be super helpful. Are there any visual references that would point to that?
2) I'm assuming since the bit-rate is higher in LB but the quality is lower, that the compression algorythm doesn't take into account potential capacity of image information? Are there intermediate codecs, visually lossless or lossless, that don't require a 4.5x bloat?