Thread: A better way to grade Ursa Mini CinemaDNGs

Page 2 of 4 FirstFirst 1234 LastLast
Results 11 to 20 of 37
  1. #11  
    Member Juan Melara's Avatar
    Join Date
    Dec 2012
    Location
    Australia
    Posts
    46
    Quote Originally Posted by CaptainHook View Post

    The 4.6K film curve does handle the full dynamic range of the sensor, but for ISO1600 for example we store into super whites to make the most of available bit depth. You'll find some other log curves are more compressed and roll off more, but they use less code values to store the range and will max at 940 or even lower in some cases. They still can cover the full range of their sensor, but they are using less bit depth than they could to store this.
    Could you explain the reasoning behind this? Decoding into 32bit float negates the need to make the most of the available bit depth by filling all the code values, as Its 32bit float! If you were decoding the image into an 8bit environment or an 8bit container it would make perfect sense. In a 10bit container you would get more precision, but unlikely you would see it in the final grade. 12bits plus, it's unlikely you would see any difference at all. So unless you transcode and work in 8bit, your throwing away Cineon compliance to solve a problem that doesn't exist in professional workflows. Or am I missing some other reason you need to fill all the code values?

    It reminds me of what Red was trying to do in 2007 with their RedLog curve, prior to implementing RedLogFilm. Part of what it did was fill up the values a lot more compared to RedLogFilm, but at the cost of being Cineon compliant. The secret sauce post houses used was the exact technique I use here, decode to linear > transform to standard Cineon. Enough post people got into their ear and RedLogFilm was developed, with full Cineon compliance.

    Since putting this out there several colourists have got in touch either confirming how much nicer LogC (or other Cineon compliant curve) behaves with the Resolve toolset. Or to tell me they had already come to a similar conclusion, and they're bypassing BMD Film on BMD cameras or others shooting CDNGs.

    Even DJI recommends bypassing the BMD Film Curve and outputting Linear > transforming into LogC using the VFX IO Linear to Arri LogC LUT. https://dl.djicdn.com/downloads/insp...03_English.pdf As it helps to better "comverge" with other cameras.

    Quote Originally Posted by CaptainHook View Post
    FYI your steps changing the RAW tab to 709 and linear are somewhat redundant, as you could just set the CST plugin to BMDFilm 4.6K for input (in the case of this old footage, don't use the V3 version) and transform to 4.6K Film V3 (and logC gamma if you prefer) from there.
    Like I mentioned in the video you can try this, but it still leaves the strange desaturation issue. The only way around it was through using RCM, ACES or the technique I show.

    Quote Originally Posted by CaptainHook View Post
    Something to be mindful of though is that the CST plugin will assume a white point of 6000K, so transforms starting from different white points will not be as accurate.
    Are you sure its 6000k and not D65? As pretty much every colourspace standard is D65 unless it specifically says so.

    Quote Originally Posted by CaptainHook View Post
    Be careful with LUTs, they often expect a specific colour space as input and may shift colours unpredictably and often further from your intention than not using the LUT - not to mention the high potential for clipping colour if the LUT assumes a wider gamut than the input actually provides. "Any LUT or process that expects log as an input" is too generic to even describe most LUTs.
    Totally agree with this and I should have been more specific.
    Reply With Quote  
     

  2. #12  
    Member Juan Melara's Avatar
    Join Date
    Dec 2012
    Location
    Australia
    Posts
    46
    Quote Originally Posted by John Brawley View Post
    Perhaps we're talking about different things.

    The LogC LUT certainly seems to do a lot more than only change a gamma curve, but I'm not THAT familiar with how Resolve works in the pipeline of processing you're talking about which is doing a transform by other means. If you're truly talking about an actual gamma curve only ? I understood a LUT could also be doing colour transforms and indeed that's what I see going on.

    The LogC REC 709 LUT is more than just a gamma curve as I understand, and have seen in end results when applied to cameras that aren't generating Alexa LogC shots.

    JB
    Actually, what you're saying is 100% correct. The LogC to Rec709 LUT is more than a gamma conversion as it also has AWG to Rec709 matrix. But I'm not using the LUT. Like Ryanite says, I'm only converting the gamma into LogC. No changes to colourspace.
    Reply With Quote  
     

  3. #13  
    Senior Member AlmaVideo's Avatar
    Join Date
    Jan 2013
    Location
    San Anselmo, CA
    Posts
    151
    Quote Originally Posted by Juan Melara View Post

    Using the CST - We manually select to debayer the raw into linear and into the Rec709 or P3 D60 colour space. The CST internally converts the input gamma into linear and converts the colour space into XYZ. Except since we are feeding it linear, only the colour space undergoes a transform into XYZ. Then we make our output choices. So in the example in the tutorial linear is converted into Arri LogC and the XYZ colour space is transformed into BMD 4.6k Film V3.
    Thanks for the tut and discussion! This kind of stuff makes visiting this user group more attractive to someone who is still very much learning.

    I don't understand why debayering into rec 709, then converting into XYZ doesn't throw away colors since rec 709 is smaller than XYZ.

    Can you explain how that works? Thanks!
    Reply With Quote  
     

  4. #14  
    Senior Member shijan's Avatar
    Join Date
    Apr 2012
    Location
    Odesa, UA
    Posts
    835
    Just done some tests. This method is kind of useless because it just produce exact same result as Davinci Color Management or Color Space Transform Node.



    Last edited by shijan; 11-23-2017 at 10:31 PM.
    Reply With Quote  
     

  5. #15  
    Senior Member CaptainHook's Avatar
    Join Date
    Aug 2012
    Location
    Melbourne, Australia
    Posts
    2,161
    Quote Originally Posted by Juan Melara View Post
    In a 10bit container you would get more precision, but unlikely you would see it in the final grade. 12bits plus, it's unlikely you would see any difference at all.
    I'm not in the Resolve team so my perspective is from the camera side and storing the data - where our SDI outputs are 10bit 422, and ProRes 422HQ (10bit) for recording is overwhelmingly the most commonly used format by people shooting our cameras (from conversations we've had with customers and even seeing recent facebook group polls etc). So for our cameras this is the reality we work with. Speaking with Harald Brendel (Arri's lead colour scientist responsible for the Alexa colour science etc) when I was at Arri HQ in Munich he says 12bit in his opinion is the minimum for LogC. This works out good with Alexa since the most commonly used format I see with Alexa is ProRes 2K/HD 4444 (12bit). We feel 10bit minimum is good for BMDFilm based on how we've optimized the curves. Something to consider here though again is Video/Data levels. 4444 is historically interpreted as RGB data range. 422HQ interpreted as YCbCr video range. Our 18% grey for all current BMDFilm log curves is 400 in 10bit (similar to logC) so you could take our ProRes 422HQ files and change to "data levels" on the clip attributes in Resolve and the pedestal/black offset and white point may be more where you are accustomed to and may look more similar to Arri 4444 prores files which are interpreted as data (cDNGs are always interpreted as Video).

    On Alexa you can actually change the SDI output from legal to extended (video/data) but again you have to be careful with workflows.
    Scroll down this lengthy post to see how on-set grading for example can be affected by this:
    http://shootdatapost.com/blog/2012/8...-extended.html

    You mention throwing away Cineon compliance - there is no such thing for camera manufacturers. Our available cameras vary in dynamic range by 3-4 stops, and our log curves are optimized to store that dynamic range in the most available bit depth. We're also not the only ones that go into the extended range to do this. Canon Clog, and Sony Slog1/2 do as well off the top of my head. Red did provide RedlogFilm but I'm pretty sure it was when they still offered no internal ProRes recording and if you're just dealing with RAW (12bit log or 16bit linear) it's less of a concern. But you'll notice they've also stepped away again from "Cineon compliance" to their new LOG3G10 curve which they are promoting as the best option - one could again assume their priority is on optimizing how they store the data.

    Quote Originally Posted by Juan Melara View Post
    Since putting this out there several colourists have got in touch either confirming how much nicer LogC (or other Cineon compliant curve) behaves with the Resolve toolset.
    This comes back to another discussion on here recently, in which John shared his experience of high end colourists and the top facilities and how often many of them have workflows/toolsets tailored to LogC/Cineon and are either unwilling or don't have the time to learn how other cameras respond, so try to force other cameras into that workflow. It's not just our cameras. Funnily enough, this is more common in the high end. But he's also found that after they do spend the time, they adjust and like it and sometimes even prefer it. This relates to both colour and gamma. We have even seen in the more high end facilities workflows reliant on dpx conversions from RAW, and colourists reluctance to go back to the RAW files to make adjustments - even when far more suited to do so. There are also ones out there pushing the edge and willing to adapt, but they seem more rare at this point in the more established facilities.

    Quote Originally Posted by John Brawley View Post
    I have worked with many senior colourists from the bigger end of town. My last show was Light Iron and the current show is CO3 with the big man himself.

    What I commonly see with colourists is an initial struggle to deal with the material, typically because many colourists have a set of pre-baked recipes and LUTs that they use for speed and process. They often struggle to match the shots to scenes that have Alexa and I can usually pick those shots.

    Then, usually by the second or third episode, they start nailing those grades, now I can’t pick the shots in the scene. The only way I can is because the Ursa is usually sharper.

    Then by ep5 I start hearing that they are liking the way the shots grade and sit. Eventually they get scenes I’ve shot entirely with the 4.6k and all of a sudden they’re not beholden to making it look like another camera. And they start to say they like it.

    On my last show, the colourist started saying he PREFFERED Ursa in low light / night shots.

    This is my experience working with many different post facilities, many colourists, in many countries
    I have spent many many hours testing the log curves from all manufacturers and in my opinion, they are all fine and handle well once you learn them and understand how to work with them. In the same way I would hope/expect a camera assist/operator/dop to do their due diligence in learning a new camera before shooting with it (when possible), I hope/expect the same from other colourists - to learn how the cameras react and how to get the best from them - not just fit them into an old workflow they don't want to adapt or change from. Younger colourists coming up tend to be more receptive to this adaptation and change, which is a common observation people have in many areas of life. Again, not everyone or all facilities are like this but it's more common than we would hope.

    Quote Originally Posted by Juan Melara View Post
    Like I mentioned in the video you can try this, but it still leaves the strange desaturation issue. The only way around it was through using RCM, ACES or the technique I show.
    Sounds like a Resolve bug with the highlight recovery and the old colour science of the 4.6K. Probably unnoticed/unreported because that colour science was only around for 2 months before we updated it with V3 being around now for a year and a half.

    Quote Originally Posted by Juan Melara View Post
    Are you sure its 6000k and not D65? As pretty much every colourspace standard is D65 unless it specifically says so.
    I'm speaking of the white point of the source illuminant, not the colour space. Using Arri again as reference, take a look at the IDT folder for exposure index 800 and notice how many files there are:

    https://github.com/ampas/aces-dev/tr...alexa/v3/EI800

    Most of these are for different CCTs (correlated colour temperature) because when doing a regression fit of a sensors spectral response to training data under different illuminants (with chromatic adaptation), different matrices are required to reduce error. So to transform Alexas Wide Gamut (the colourspace has a D65 whitepoint, yes) to ACES primaries (roughly D60) you also need a different matrix for each CCT for the most accurate conversion (they also have IDTs for when their internal NDs are used). With cDNG the Adobe approach is to have two matrices, one for Illuminant A (2856K) and the other daylight, then interpolate for CCTs in-between which works quite well also.
    Blackmagic Design
    My BMD LUTs.

    **Any post by me prior to Aug 2014 was before i started working for Blackmagic**
    Reply With Quote  
     

  6. #16  
    Senior Member
    Join Date
    Apr 2016
    Posts
    142
    I hope/expect the same from other colourists - to learn how the cameras react and how to get the best from them - not just fit them into an old workflow they don't want to adapt or change from. Younger colourists coming up tend to be more receptive to this adaptation and change, which is a common observation people have in many areas of life. Again, not everyone or all facilities are like this but it's more common than we would hope.
    a couple of thoughts on this one ;-)
    - pipleine facilities are constrianed by budgets / scheduales / post supervisor's comfort zones / bond agency's dislike of anything not proven , and do turn to a well known pipeline, does not mean the colorst prefers this, only that the facility mandates a workflow to be used in consultation with many parties involved, and we ahve to work within the boundries laid out, or start our won facilities and find our own clients...
    btw that's the route i've taken although i still freelance at large facilities at times, usualy on features, three films this year in facilities with a Christe + X300

    alot of the mot tech savy and cutting edge colorists have engineering backgrounds and do understand the maths underneath the image transforms, a bit more than just knob twiddlers

    And all that said, i use ACES in 99.9% of my work, so much simplier and fewer / cleaner transforms - just finished a series shot on the 4.6, and everyone was very happy with the end resualt, no flys on that guy, but also i see no need to mess around with it, jsut keep it simple, clean and let it be what it wants to be, but then i take the same approach with RED or Alexa shows ;-)
    Reply With Quote  
     

  7. #17  
    Senior Member
    Join Date
    Mar 2017
    Location
    Melbourne, Australia
    Posts
    613
    Quote Originally Posted by dermot shane View Post
    a couple of thoughts on this one ;-)
    - pipleine facilities are constrianed by budgets / scheduales / post supervisor's comfort zones / bond agency's dislike of anything not proven , and do turn to a well known pipeline, does not mean the colorst prefers this, only that the facility mandates a workflow to be used in consultation with many parties involved, and we ahve to work within the boundries laid out, or start our won facilities and find our own clients...
    btw that's the route i've taken although i still freelance at large facilities at times, usualy on features, three films this year in facilities with a Christe + X300

    alot of the mot tech savy and cutting edge colorists have engineering backgrounds and do understand the maths underneath the image transforms, a bit more than just knob twiddlers

    And all that said, i use ACES in 99.9% of my work, so much simplier and fewer / cleaner transforms - just finished a series shot on the 4.6, and everyone was very happy with the end resualt, no flys on that guy, but also i see no need to mess around with it, jsut keep it simple, clean and let it be what it wants to be, but then i take the same approach with RED or Alexa shows ;-)
    Oh wow I don't know why I've never heard of this before, looked up ACES and a couple tutorials. So simple and so good, got my image to such a nice place so quickly and I really like that it's some sort of industry standard. Really good to learn, thank you for the mention! otherwise I'd never have found it.

    this is the blog I was reading for anyone wanting to learn it. https://www.premiumbeat.com/blog/wha...color-grading/
    Reply With Quote  
     

  8. #18  
    Moderator
    Join Date
    Apr 2012
    Location
    Atlanta Georgia
    Posts
    2,916
    Quote Originally Posted by dermot shane View Post
    - pipleine facilities are constrianed by budgets / scheduales / post supervisor's comfort zones / bond agency's dislike of anything not proven , and do turn to a well known pipeline, does not mean the colorst prefers this, only that the facility mandates a workflow to be used in consultation with many parties involved, and we ahve to work within the boundries laid out, or start our won facilities and find our own clients...
    But I've never seen a workflow document the specifics of how the workflow works within a given platform. At most I've seen a platform specified (ie Colorfront Express Dailies for dailies and Resolve for final finish)

    Nowhere have I seen what LUT's or CST's are used. That's always up to the colourist.

    I understand why certain tools are used, but as Hook says, there's a certain stubbornness to some colourists about trying something else and stepping out their workflow comfort zone. I get it. I'm the same with the way I like to light and shoot. But we have to be open to learning and trying other things. I've already learned from this thread that there's a difference between a Arri 709 LUT and a LogC CST (gamm only) transform.

    What about the fact that there's certain workflows that assume shooting at 6000K ? I've never shot that white balance ever, and yet, there's a whole workflow that's hanging off that little known assumption.

    I am no expert on post, but it often feels like I know a heck of a lot more than most of the post supervisors that I meet, not to mention others at the post facility itself. I wish it were the other way and it should be.

    JB
    Reply With Quote  
     

  9. #19  
    Senior Member
    Join Date
    Apr 2016
    Posts
    142
    But I've never seen a workflow document the specifics of how the workflow works within a given platform. At most I've seen a platform specified (ie Colorfront Express Dailies for dailies and Resolve for final finish)
    i have emails from major facilities stating RCM, Truelight & ACES are not to be used on the film i'm hired to grade, only display refered workflow, only useing flatened EXR /DPX/ProRez, no access to the raw /cam orig...
    you may not be aware of what we are asked to take on, but that only means you are not aware of what's going on in the background of a post house, lack of awareness is to be expected, post houses do not share this stuff with cleints generaly


    there's a certain stubbornness to some colourists about trying something else and stepping out their workflow comfort zone.
    i know more than a few colorists, don't know of anyone like that tho....


    What about the fact that there's certain workflows that assume shooting at 6000K ? I've never shot that white balance ever, and yet, there's a whole workflow that's hanging off that little known assumption.
    i have zero clue what youre on about here, can you be even a tiny bit more specfic?

    I am no expert on post, but it often feels like I know a heck of a lot more than most of the post supervisors that I meet, not to mention others at the post facility itself. I wish it were the other way and it should be.
    no matter what you think of the post sup, and how highly you rate your skills in post, we have to respect the post sup's call's, just as we respect the DP, the Director, the Producer's, the QC houses, and the bond agency...

    overall i thibk you and Mr Hook may be over rateing yourselves in terms f knoledge of post where the rubber meets the road, but then you have said you are not a colorist, so you get a partial pass, opnions with out a solid foundation and based on a lack of knoledge are only that... opnions
    Last edited by dermot shane; 11-24-2017 at 09:51 PM.
    Reply With Quote  
     

  10. #20  
    Senior Member
    Join Date
    Mar 2017
    Location
    Melbourne, Australia
    Posts
    613
    Quote Originally Posted by dermot shane View Post
    i have emails from major facilities stating RCM, Truelight & ACES are not to be used on the film i'm hired to grade, only display refered workflow, only useing flatened EXR /DPX/ProRez, no access to the raw /cam orig...
    Why would they not want you using ACES and how often is this the case? I remember reading a part of the tutorial where you render without ACES on and it gives out a trippy image but allows others to view in their preferred display method, then you give them a lut. That was said to be for hand offs/ vfx etc. Is that the same as what you are talking about or?
    Reply With Quote  
     

Similar Threads

  1. Replies: 4
    Last Post: 06-09-2016, 04:53 PM
  2. Grade My DNG 4.6K Mini
    By Abstrak in forum General Discussion
    Replies: 10
    Last Post: 03-28-2016, 04:06 PM
  3. Grade JB's Mini 4.6k
    By jambredz in forum General Discussion
    Replies: 128
    Last Post: 03-26-2016, 01:49 PM
  4. Replies: 7
    Last Post: 07-09-2015, 03:48 AM
  5. Finally, some CinemaDNGs of the Digital Bolex
    By Abobakr in forum Off Topic
    Replies: 5
    Last Post: 08-11-2013, 11:14 PM
Bookmarks
Bookmarks
Posting Permissions
  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •