Thread: FULL FRAME TEST MONSTRO/ALEXA LF/MAVO/VENICE

Reply to Thread
Page 2 of 9 FirstFirst 123456 ... LastLast
Results 11 to 20 of 83
  1. #11  
    Senior Member Aaron Lochert's Avatar
    Join Date
    Oct 2013
    Location
    Tucson, AZ
    Posts
    918
    Quote Originally Posted by rand thompson View Post
    Here is the After Recovery. Some of the Whites where not recoverable such as the light to his right, the playing cards to his upper-right and the Coffee mug above him. Plus there is some Magenta in the greyish walls. Also the Colors of both of the color charts look really funky. However, the skin tones where recoverable to a point, But I don't think I would still ever use this image.
    Let's be honest with ourselves, though. When brought into your grading app of choice with IPP2 and white balanced via raw controls (AKA, the intuitive thing anyone would do), we're going to get this image:

    800 ISO, 5450k, -5.5 tint, -5.7exp IPP2 M/M


    And that's with the current highlight extension algorithm.

    In order to recover better than that, we have to do a very counter-intuitive thing and set the raw white balance away from white (this one is set to 3800K), and then regrade the color back:



    And I've found that dialing the color back to neutral is a bit involved and often requires some RGB curves. Luckily in this example we have a greyscale chart to work with, but you won't be so lucky in many scenarios. Here's the RGB curve that was needed to neutralize the offset from the 3800K setting.

    So while there is some recovery you can do, it's hidden away in hard to reach places that many won't find. The few of us know about it from perusing these threads, but it doesn't surprise me that many non-techy people are going to miss it. Also, we should not think of these recovered stops in the same regard as unclipped stops on other cameras. Graemme said it best: "It would be unwise to rely upon clipped regions for shooting. Best to think of them as gravy rather than the main course."
    Reply With Quote  
     

  2. #12  
    It's a good test but you are relying on them working with each camera 'to the best of it's ability'

    I think Aaron above points out something really valid - there are approaches, scene dependent, for getting the best results. That's an interesting approach, but it doesn't make sense to me why it would work. So i'm going to have to do it myself to understand. Is that R3D downloadable?

    In the test you can see how Arri favours the highlights and i thought the Mavo did *really* well all things considered.

    cheers
    Paul
    Reply With Quote  
     

  3. #13  
    Senior Member rand thompson's Avatar
    Join Date
    Aug 2011
    Posts
    4,171
    As stated above, Yes, this is a counter-intuitive method to try to recover more than would practically be achievable. However, I thought that was fairly obvious and didn't need to be explained. The point was to show that there were other than the usual methodology to get better than usual results from an IPP2 derived image.

    No one that I know of would use any of the test images from this page or the previous page, hopefully. But we have to be carefully about what are defined as conventional and non conventional methods of solving problems. I don't use RGB curves , Power Windows, nor do I use qualifiers anymore , I just don't like what they do to an image, IMHO. However, most people that would use Davinci Resolve would use exactly those tools.So who am I to say that their methods are better or worst than mine.
    Reply With Quote  
     

  4. #14  
    Quote Originally Posted by rand thompson View Post
    As stated above, Yes, this is a counter-intuitive method to try to recover more than would practically be achievable. However, I thought that was fairly obvious and didn't need to be explained. The point was to show that there were other than the usual methodology to get better than usual results from an IPP2 derived image.
    My curiosity is because the math doesn't really make sense unless the workflow is messed up.

    The R3D files hold the data in camera space and each channel has its clipping point, that's finite.

    When it's debayered, usually into a huge agnostic space like XYZ (this may or may not happen to red files). Then the precision is such that no matter how you choose to white balance, you should not see clipping for one white point and not another. Highlight recovery works in this agnostic space as well.

    I'm basing the knowledge on how i've seen and used more open RAW platforms like cDNG so the R3D pipeline could be different.

    It is a friday and it's been a long week so i might be just missing something obvious here. Hence wanting to try it out for myself.

    cheers
    Paul
    Reply With Quote  
     

  5. #15  
    Senior Member rand thompson's Avatar
    Join Date
    Aug 2011
    Posts
    4,171
    Paul,

    I think I know what you're saying. If something is clipped no matter how creatively you white balance that image it should still be clipped

    That was my understanding as well, until I started pushing and pulling R3d files until they almost and finally did explode. There seems to be maybe some untapped potential in the IPP2 workflow or atleast the Color Science behind it.

    It's funny that we are discussing conventional and unconventional image development workflows in the same breath as IPP2, which in itself is one of if not the most unconventional way of bringing a Log image into a REC 709 colorspace than of any of the other manufacturers. The log to REC 709 transform lut has to be applied at the end instead of the beginning as with other manufacturers. Creative grading and luts have to be applied before the log to REC 709 transform lut instead of afterwards. And qualifying is somewhat more difficult than other Log transferred images.

    I think this is why a lot people who are use to more traditional color grading workflows are still trying to wrap their brains around it. I'm using ACEScct more these days and that has it's own percularities.
    Reply With Quote  
     

  6. #16  
    Quote Originally Posted by rand thompson View Post
    Paul,

    I think I know what you're saying. If something is clipped no matter how creatively you white balance that image it should still be clipped
    Sort of. in the example above the recovered face shows some detail not in the first version. So either that detail was recovered (aka made up) using highlight recovery (which basically means using unclipped channels to rebuild clipped ones) or there is an actual an issue going on in that in some cases highlight data is being clipped in a particular processing pipeline.

    This seems odd to me so i was wondering where that frame came from and whether i could get an R3D to see for myself and try in a few different apps to understand what's happening.

    EDIT - sorry, i'm being dense, i see the link in the post above - my bad...

    Quote Originally Posted by rand thompson View Post
    That was my understanding as well, until I started pushing and pulling R3d files until they almost and finally did explode. There seems to be maybe some untapped potential in the IPP2 workflow or atleast the Color Science behind it.
    Or if you can show me another example that would be great!

    Quote Originally Posted by rand thompson View Post
    I think this is why a lot people who are use to more traditional color grading workflows are still trying to wrap their brains around it. I'm using ACEScct more these days and that has it's own percularities.
    I dunno. In Resolve i use LogG10/RWG and use the tone mapping from Red to move into 709 or P3. I find in super saturated cases it gamut maps with a better understand of the camera. But if you have the end of a pipeline using ACES then that's fine, so long as your IDT is correct.

    cheers
    Paul
    Reply With Quote  
     

  7. #17  
    Senior Member rand thompson's Avatar
    Join Date
    Aug 2011
    Posts
    4,171
    Paul,


    here is the link


    https://drive.google.com/drive/u/0/f...kaCeokqlam6rON





    Just "Right-Click" on the "+5 Standard ISO". It will create a Zip file and after its finish it will ask you where to save it to



    Reply With Quote  
     

  8. #18  
    Senior Member rand thompson's Avatar
    Join Date
    Aug 2011
    Posts
    4,171
    Paul,


    I don't own a Red camera , but maybe someone will provide some +5, +6 ,+7 stops overexposed .R3ds for us to play with, HaHa.
    Reply With Quote  
     

  9. #19  
    Senior Member
    Join Date
    Nov 2008
    Location
    Dallas, TX
    Posts
    2,760
    Wow this test is pretty cool. I love the big layout, very revealing. I don't think I've ever seen it presented quite this way. It is truly incredible how old the arri sensor is and how they just nailed it dynamic range and color-wise.
    JAKE WILGANOWSKI
    Director of Photography / Filmmaker
    CINE-AUTOMATIC.COM
    Reply With Quote  
     

  10. #20  
    Senior Member Aaron Lochert's Avatar
    Join Date
    Oct 2013
    Location
    Tucson, AZ
    Posts
    918
    Quote Originally Posted by rand thompson View Post
    but maybe someone will provide some +5, +6 ,+7 stops overexposed .R3ds for us to play with, HaHa.
    Here's Helium 0-6 stops over, including white balanced and exposure corrected RMDs: https://drive.google.com/open?id=1_a...TI8J51VlCW9IKB

    But let's not derail this thread any further since this isn't specific to full frame sensors. Perhaps it's time to revive this thread.

    Edit: Forgot to mention lens was at an f/2.0
    Last edited by Aaron Lochert; 06-07-2019 at 09:56 PM.
    Reply With Quote  
     

Posting Permissions
  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts