
- Posts: 9535
- Joined: Tue Sep 11, 2012 10:20 am
- Location: Poland
Why ProRes4444(XQ) out or Resolve is not RGB based, but YUV 4:4:4?
It would be better to have it as RGB (or ideally have option to choose) as then there is way less constant RGB<->YUV conversion further in the production chain. As a very high quality codec and often used not for final delivery, but as intermediate production format it should be RGB based.
Apple quite clearly says it can be RGB:
"Direct encoding of, and decoding to, RGB pixel formats"
which is exactly what we want, instead of constantly going YUV<->RGB on every import/exports.
(https://documentation.apple.com/en/fina ... tasks=true)
There seems to be some contradictory statements as quite many industry people keep saying ProRes is alway YUV based, yet Apple in white paper is clearly saying RGB based encoding is also supported. Yet, I have never seen RGB based PorRes file (if mediainfo is to be trusted).
Resolve ProRes444 is actually "strange" as it uses src_pix_fmt=14, which is "invalid". This field suggest what uncompressed data was fed into code for encoding. Possible options are here:
https://wiki.multimedia.cx/index.php/Apple_ProRes
Most typical for 4:2:2 option is v210.
For 12bit RGB we should expect 'b64a' (16-bit ARGB). 12bit YUV 4:4:4 should go over 'r4fl' (32-bit floating-point 4:4:4:4).
In contradiction to ProRes444, DNxHR 444 is exported as RGB, which is good. It also perfectly supports 12bit precision. In case of ProRes4444 there is some "strange" noise in last bits, so it's more like 11.5 bit, not 12. This may be related to actual fact that data is YUV encoded and has to be converted to RGB (in Resolve or AE).
Another hint that ProRes4444 in Resolve is YUV based is fact that new super white/black preservation option is enabled, where for DNxHR 444 is not (as it's RGB based).
I would really like to get some official statement about it (and option to be able to choose RGB/YUV for DNxHR/ProRes 444).
It would be better to have it as RGB (or ideally have option to choose) as then there is way less constant RGB<->YUV conversion further in the production chain. As a very high quality codec and often used not for final delivery, but as intermediate production format it should be RGB based.
Apple quite clearly says it can be RGB:
"Direct encoding of, and decoding to, RGB pixel formats"
which is exactly what we want, instead of constantly going YUV<->RGB on every import/exports.
(https://documentation.apple.com/en/fina ... tasks=true)
There seems to be some contradictory statements as quite many industry people keep saying ProRes is alway YUV based, yet Apple in white paper is clearly saying RGB based encoding is also supported. Yet, I have never seen RGB based PorRes file (if mediainfo is to be trusted).
Resolve ProRes444 is actually "strange" as it uses src_pix_fmt=14, which is "invalid". This field suggest what uncompressed data was fed into code for encoding. Possible options are here:
https://wiki.multimedia.cx/index.php/Apple_ProRes
Most typical for 4:2:2 option is v210.
For 12bit RGB we should expect 'b64a' (16-bit ARGB). 12bit YUV 4:4:4 should go over 'r4fl' (32-bit floating-point 4:4:4:4).
In contradiction to ProRes444, DNxHR 444 is exported as RGB, which is good. It also perfectly supports 12bit precision. In case of ProRes4444 there is some "strange" noise in last bits, so it's more like 11.5 bit, not 12. This may be related to actual fact that data is YUV encoded and has to be converted to RGB (in Resolve or AE).
Another hint that ProRes4444 in Resolve is YUV based is fact that new super white/black preservation option is enabled, where for DNxHR 444 is not (as it's RGB based).
I would really like to get some official statement about it (and option to be able to choose RGB/YUV for DNxHR/ProRes 444).