Prores 4444 in resolve

Dominik Bauch Oct 18, 2016

  1. So in resolve I'm mixing Arriraw & prores 4444, both shot on alexa. If I export a graded arriraw from the timeline as a prores 422 HQ QT then all looks as expected.
    But if I export a prores 4444 clip from the timeline as the same prores 422 flavor as above then it comes out lifted and washed out.
    Any ideas what should be done? I know 4444 is a data levelscodec, so I set the clip attribute data level to Full, but having this at auto or full seems to make no difference. It looks normal on the timeline when next to the other arriraw clips from the same scene.
    Any ideas?
  2. ARRI uses legal range for all their ProRes formats including ProRes 4444. Resolve is misinterpreting it as full range.

    From ARRI website:
  3. I have never seen two renders from the exact same session come out different. What happens if you render out a test signal like bars in 422HQ vs. 444? What are you watching the files on? In Video Monitoring, are you running the session in "Data Levels - Full" or "Data Levels - Video"?
  4. can you find a white paper that confirm this?
  5. Erik Wittbusch likes this.
  6. As another curiosity, I was working on a project, where I used Resolve for pre-conform. Timeline was a mix of Alexa Prores and Phantom. The reason for pre-conformed workflow there were high speed Phantom images, that were shot at night, hence very noisy images. Neat Video needed to be used on almost every Phantom shot. I had no problem rendering HD timeline on Mac, but for UHD, Mac ran out of juice with Neat Video. I have GTX1080 installed in my HP computer, so I opened the project there and I was easily able to render it in UHD. The only problem it was on Windows, so DNxHR444 was used. I brought it back into a Resolve and reapplied all my grades from HD timeline, only to find out, that nothing really matched, especially black levels. Yes, I know, there are many variables in my story and I expected a little difference, but I was really taken back with huge difference between Prores 444 and DNxHR444.
    Very strange...
  7. How did bars look? Could this be just a simple Video / Data Levels issue, or is there something more dramatic going on, like a colorspace/color management issue?
  8. Both were done using video levels. But even then, applying a global trim didn't really work. Some shots were just different in some other ways...
  9. The increase in these questions really reenforces the need to scientifically test the whole workflow in pre-production. Marc is so right that you need to run bars carefully through the production path and check and reconfirm black levels accurately as it is really easy to miss.
    This should be the basis of codec, camera and software choices, rather than, the cinematographer just doesn't like Pro Res, or we can't afford RAW etc.
    The good thing about this is that more people are becoming aware of the situation and problem, that is often invisible when the players and monitors hide the conversion from you.
    I was under the impression that 444 was always full range, so you really need to do a lot of research and testing to maintain signal integrity.
    In fact, forget research. Don't trust anything anyone ever says. Test it with black and white levels from video to full range and double check them carefully with scopes.
    Argue emphatically for the budget to this, and actually do it. I'm looking at you zombie DITS!
  10. I wonder if it possibly could be a 2.2 gamma vs. 2.4 gamma issue. Something is definitely strange going on. What I can't figure out with Jake's issue is whether every shot was equally wrong, or if some shots were more different than others.
  11. If you read my response again, I already have said:
    "Both were done using video levels. But even then, applying a global trim didn't really work. Some shots were just different in some other ways..."
  12. Well, I ask again: how do test signals look? What happens with a second of SMPTE bars and a second of grayscale ramp at the head of program? Heck, do the same at the end. Is it possible there's a timeline grade involved? The only other possible cause would be a color management issue, and there's hundreds of potential combinations there.

Share This Page