Prores 4444 renders look different (individual source clips vs single clip)

[Joseph Owens]

[Evan Kultangwatana] “when using prores4444, does resolve waffle between YUV and RGB, or different gammas depending on your render settings?”

As a rule, no, it does not. It pretty much does exactly what you ask it to. But other playback applications, Final Cut, Quicktime,… etc., which are also subject to the System ColorSync preferences, do.

I am supposed to be generating an article about this, with diagrams containing circles and arrows, but what you are encountering is ambiguity between applications in how RGB / Y’CbCr scaling is handled in different codecs that are capable of carrying either value map. In the simplest case, “full scale” RGB in 10-bit is regarded as 1-1023. Linear video is scaled 64-940. Obviously if one is misinterpreted as the other, the display values will be noticeably different. Unfortunately, codecs like ProRes444 don’t have a flag to indicate what they are carrying, so you either have to tell the application what it is (as Media Composer requests) or hope that the display app “guesses right.” They don’t, usually.

Use of an outboard scope makes this obvious, because the outputs are measured independently, displayed as numerical values, and you can see that the black and white points are not being placed correctly. This can manifest as the “gamma bug”, because most people operating simple systems are eyeballing their media and the relative contrast/brightness is similar to that defect — but this should all be in the past, if your operating system is later than OSX10.6.8., because Apple standardized (finally) on 2.2-2.4 rather than 1.8 with that version.

jPo

“I always pass on free advice — its never of any use to me” Oscar Wilde.

[Evan Kultangwatana]

I understand all the ins and outs and pitfalls of gamma, codecs, etc.. But I’m fairly certain that what I’m seeing is an outright bug. Because I’m using the same exact render settings, but getting different results depending on whether I’m doing a full render or individual clip render. It’s a slight difference, but on this particular job critical details really matter.

This is a nMP, D500 machine. Resolve 11.1.3 on OSX 10.9.5.

[Joseph Owens]

[Evan Kultangwatana] “using the same exact render settings, but getting different results depending on whether I’m doing a full render or individual clip render.”

I can see where this might happen if there was some other filter applied to the clip that would affect its output behaviour with respect to something like transparency or some other layer-related hierarchy. Clip renders would tend to ignore transparency, for example, and re-apply that in the NLE upon round-trip re-entry, as opposed to single clip, which assumes that the output is basically a final pass and observes factors like composite assignment.

jPo

“I always pass on free advice — its never of any use to me” Oscar Wilde.

[Marc Wielage]

[Evan Kultangwatana] “Because I’m using the same exact render settings, but getting different results depending on whether I’m doing a full render or individual clip render.”

Render out some SMPTE bars and a grayscale test pattern and see how it looks on scopes. This will tell you the truth and possibly help you pin down a cause and a solution.

Maybe I’m paranoid, but I always, always include a few frames of bars and grayscale with all my renders, just so if I think I’m seeing a problem, 5 seconds later I can check the scopes and say, “nope, no problem” or “whoa! We have a problem!”