Dan Hudgins on the KineMINI Workflow
August 31, 2014 at 11:57 am #705
The monitoring in the camera is mirrored in a 3D-LUT saved in each shot folder, using that can give you results that look at least somewhat like what you saw on the camera’s monitor.
Since the camera is raw recording, linear 12bit to the DNG and log encoded non ISO normalized to the Cineform raw, you can re-grade any curve and get the full range of ISO curves back from the raw data if you don’t like what the monitoring mirror gives.
They say they are working on KineSTATION to give some extra options for RGB output that will be able to mirror the monitoring at the time of shooting, and maybe also add the option for 3D-LUT user look in the camera that would be added to the mirror 3D-LUT in the shot folder, but those things are not finished yet.
For now the 3D-LUT in the shot folder works with Speed Grade (when the supplied 1D-LUT also in the shot folder is used to convert the linear DNG into log90 same as the Cineform raw is encoded) and with programs that support the Cineform raw codec and you copy the 3D-LUT’s from the shot folders into the Cineform raw codec, manually.
If you view the Cineform raw you will see the log90 encoding WITHOUT the correct matrix applied if you do not copy the shot folder 3D-LUT into the CiIneform codec LUT management, that is not KineLOG and has nothing to do with the KineLOG look group in the camera’s setup menu, for some reason they are confusing two things that are very different by using the same KineLOG name to describe them,
1) KineLOG look group is ISO normalized and has matrix applied to simulate film log film scans for making DCI-P3 for digital projection with some grading.
2) “KineLOG” out of KineSATAION is the Cineform raw log90 curve that is fixed for all ISO and does not change with ISO adjustment, and has no matrix applied so that if you just increase saturation the amount of primary colors will not be balanced since no balancing matrix was applied, the “density” will also not correspond to the exposure or ISO monitoring curve selected.
To actually get KineLOG or the other look groups Kine709 and KineCOLOR you need to apply the shot’s 3D-LUT that is taken from the monitoring settings at the time of shooting.
Hopefully the RGB output options with the monitoring mirror 3D-LUT in the shot folder will help people not able to manage working with raw data get results close to what they adjusted the camera settings to look like when shooting, this is an on-going development.
For now your options are:
1) Use Speed Grade with two of the LUT in the shot folder.
2) Use programs that work correctly with the Cineform raw codec and copy the shot folder 3D-LUT for monitoring mirror matching into the right part of the Cineform program folder, there is a signature key in each 3D-LUT and MOV file that Cineform raw codec uses to link them on playback, the codec will switch to the right 3D-LUT on-th-fly when playing and rendering if you install ALL the needed 3D-LUT into the codec manually before you play or render from the MOV clips.
3) Grade on your own from the DNG raw data section ignoring the header data (which is what I do in my de-Bayer program for my Vimeo videos).
4) Load DNG using the header data and then try to compensate for errors due to each program using its own method of finding the ISO normalization. Changes in the header are planed to improve this for use in some programs. Note that some programs use AGC and so are useless for Cinema work, others may apply fixed gain making mid-tone very depending on which ISO curve was used in the camera for monitoring. You should make sure the raw data is white balanced at the time of shooting for the matrix in the DNG header to be anywhere close to right. The DNG header matrix is not the same as the monitoring matrix due to the way Adobe does their color processing there will not be an exact match to how the camera’s real-time processing works, and some programs using their own DNG code or the Adobe SDK may clip the highlights when in fact there is highlight detail in the raw data stored in the DNG frames (which is why I wrote my own code just for cinema use, not
based on still camera snapshot processing as Adobe’s DNG SDK started out for).
The program that processed the raw Bayer data has to guess what values to replace the missing parts of the Bayer pattern with, so no two programs using different methods will show the same artifacts, in addition to that the OLPF compensation sharpen and any re-size or letter boxing can introduce aliasing independent of the cameras own issues, as all digital images no matter what their source are subject to the pixel grid pattern and resolution limits of that as well as aliasing on re-size or re-sample etc.
Although RGB images may help, to get best results the lighting should be blanced for a good highlight to shadow ratio, its not just a matter of the camera’s dynamic range, its a matter of what one can later get to look good within the Rec709 or DCI-P3 or “monitor gamma” type display limits, so good lighting is needed to get good results, as are makeup, costume design, set dressing, and over all color consultation (set paint colors, gels on windows, etc.).
I have made new look groups for testing to emulate ARRI LOG-C (KineLOGC), Sony S-Log(1), S-Log2, S-Log3, (KineLOGS1, KineLOGS2, KineLOGS3) Canon CLOG (KineLOGC), and a revised Cineon film log KineLOGF, and revised Rec709 (Kine709B).August 31, 2014 at 7:26 pm #881
You must be logged in to reply to this topic.