Color shift while exporting Apple ProRes from Premiere Pro

Hi @kazuyaf88705976 -  Thanks for submitting your bug report. We need a few more details to try to help with the issue.
Please see, How do I write a bug report?

What version of Premiere Pro are you using?

As others have said please provide reproducible steps for your issue. 

In the Lumetri Panel > Settings you may have different viewer gamma settings set so that is why your exports look different.  Here is a helpx document Why do my colors look washed out

Your colors may not look washed out, but same concepts.

Sorry for the frustration.

Hey there,

I believe what you may be experiencing is video and full levels mismatches. I forgot the exact details on where which goes wrong but Resolve definitely has issues. Inside Resolve you can actually correct any mismatches by going into the clip attributes and setting it to Video or Full instead of Auto.

Adobe will only detect from file and scale accordingly, the user does not have any control over interpretation. When you write APR444 from Adobe I believe it always writes Video levels eventhough the expected formatting for writing RGB instead of YUV is full levels.

Anyways, try a manual export out of Mac Resolve and Windows Resolve with data levels set to Full and another to Video and see what happens in both Premier and Resole on both Mac and Win to see where it goes wrong.

You didn't answer my first question, which was what export settings you used from Resolve?

Color Management has so many variables now it can take some time to nail differences.

Hello, thank you for the comment.

I understand that no two monitors display the exact same image, so I don’t mind if there are slight color or gamma shifts when viewing on different displays.

The issue I’m facing is with exported ProRes 4444 files. Since it's supposed to be a lossless codec, the RGB waveform shouldn't change at the data level. At the very least, when viewing the same footage in the same software on the same monitor, the waveform of the original Canon MXF and the exported ProRes should match. However, the waveform of the ProRes file is noticeably altered, which I believe is a problem.

I'll explain why, which takes some space. For the short skinny, how you might prefer to set things, go to the bottom of the post.

----

The specification for working with Rec.709 is extensive, and includes amazing details for working in image correction. And so to work in full specs, it includes the monitor and room brightness levels.

The specified monitor setup is sRGB primaries, D65 white point, a power-law curve essentially similar to gamma 2.4, and a screen brightness of white at 100 nits.

Then they have the viewing situation specced also. The environment is a pretty dark though not quite blackened room, I don't recall the general lux level, but it's pretty dim. Any wall or surface visible to the worker when working must be totally neutral gray, the value of the paint specified. There must be a 'bias light' between the monitor and the wall so that there is a very small lightened area of an exact amount around the monitor from the working position.

And of course, all lights in the room must be D65. So it's dark enough most colorists need a small light on their worktop ... and that's got to be a pretty darn dim D65 one. 

So working for total evaluation of the image is supposed to be done in that darkened space. And in that darkened space, the 100 nits screen is quite bright enough.

But ... working in a normally lit room, the specs for the monitor go to essentially gamma 2.2.

So for actual image verification, you need to be in a "semi-darkened room" ... funsies. And yes, my working environment is quite tight to the specs. If I go full colorist mode.

For most people, the working area is not going to be that dark, and working with the screen at gamma 2.2 is better as it visually lifts the shadows a bit for you. However, from the extensive testing that resulted in those specs, it is known that in that brighter room, it is not a good environment for color/hue judgements. Especially color saturation.

Ergo the requirement for full-on colorist work to be done in the darker room.

So this is a bit tricky to get setup at first, and colorists have always had an issue since flat-screens replaced CRTs with clients looking at the image on their screens. Most of the many I work with have specs in their contrast about even how judgements can be accepted.

Such as either from in-room work with them, or now, quite a few use specific iPad models that they own and mod the screen settings so they get fairly close visual match to their reference monitor. Their contract says change requests or orders must be done in-room or from viewing only on that iPad in a fairly dark room.

I hope that covers the problems adequately. 

Also, note, no two screens can possibly show the identical image! 

Due to the vagaries of manufacturing tolerances no two pixel elements will be exactly the same in micro amounts, let alone the colored filters that are placed over them in a pattern so that the camera can be tricked into creating color by complex math, or ... the diodes that produce color in screens. 

This is part of why Grade 1 Reference Screens (which is a very strict specification itself) are so expensive. The makers of screens measure every screen coming off the line, and the most accurate ones are sold for a vastly higher price than the ones with more variance.

Grade 1 Reference monitors are the highest, tightest tolerance; expensive computer and TV makes come next, on down the line to the cheapo monitors and TVs.

But even those spendy Reference monitors, two "identical" models side by side, calibrated with the same expensive photo spectrometers, given both a calibration and a profile pass, fed images from the same breakout device from BlackMagic or AJA, so in their profile passes the charts produced by the software are virtually identical ... and yet ...

With a high-end group of viewers, experience pros, in a darkened room ... they will see differences between those two monitors.

Now ... throw in the variables of cheaper screens, what the OS and screen makers do to "enhance the viewing experience" ... user settings ... and the variablity of the lighting that the user views things in.

And beginning colorists are told this dictum:

NO one, on any screen by any deliverable method (including network broadcast, streaming, or theatrical release) ... will ever see exactly what you saw while grading. It is not physically possible.

So why the specs? Well, if you follow those in doing your grading, then, in relative terms!!!! ... your outputs will look similar to other professionally produced media on any screen.

There's another phrase ... "You can't fix gramma's green TV." Truth.

So try this: set the Mac monitor to Reference mode HDTV, but in Premiere, use the viewing option for gamma 2.2.

This will probably give you more what you would like to see.

Will the Mac viewing transform affect exports when using Apple ProRes 4444 in Premiere Pro?
I switched to the HDTV color profile, but the screen is way too dim, and I don’t have any control over the screen brightness. I’d prefer to avoid using that mode if possible.

That sounds like the problem caused by Apple using a non-standard display transform gamma in their ColorSync utility. Which is 1.96 rather than the specified 2.4.

On the Mac, IF you have a Reference modes option, set it to HDTV and you get a correct display transform. 

If not, try VLC or Potplayer, neither of which allows ColorSync to control the image. See if it's the same in that player as Premiere. You can do that with the files you've already created that had this issue.

Hello,
I was about to prepare the photographs, but my Premiere Pro license expired yesterday, so I’m no longer able to export images.

I also noticed that footage showing no color shift in Resolve on Mac actually had a color shift when viewed in Resolve on Windows. Conversely, footage exported from Premiere on Windows didn’t show any shift when viewed on Windows.

I’ve realized the safest approach is to avoid converting my footage to ProRes until I decide which machine I’ll use to deliver the final product.

It’s probably an issue related to the combination of the graphics card and operating system—it doesn’t seem to be a problem exclusive to Premiere.

Since I don’t have a way to delete the post, I’ll leave it as is. However, I’ll revisit this topic and ask the same question again once I reactivate my account.

Shebbe has the questions we'd need to puzzle this out.

Just one more question ... which export option were you using in Resolve? In very specfic detail.

Hey @kazuyaf88705976 ,

Perhaps because you're a new user but we don't see any images in the post. Could you try sharing them again?

Furthermore, did you bring the prores file back into Premiere Pro to compare? Or were you using an external video player.

Also to help the developers reading this thread better understand the issue could you share your system info and Premiere Pro version?

Is the problem only present when exporting to ProRes 4444? Or is it in ProRes 422 flavors as well. What about other formats like H.264/5.

Lastly, what happens if you grab a still frame of a Canon MXF clip, put it back in the timeline and export that to ProRes in both Premiere and Resolve. Does that present the same data? This could help track down where the issue happens further.