7219 Red Comparison

[Dan Hudgins]

Can one size fit all?

Can one size fit all?

Do think it will be possible to create a filter or process that can be applied to RED footage generically?

The notion of light having a single K value is false for most artificial light sources, and even many daylight sources.

When an object is near other objects they reflect light onto each other, so the light picks up colors that act like bandpass filters.

When films were shot under studio lights the DP did not have to deal with the time of day and weather issued of Daylight, and did not have to deal with odd artificial light sources on location.

To do a good job of Grading one needs to look at the shot and try to find a setting that works for all the frames, since you do not want the Grading to change during a shot most of the time.

Grading is not a substitute for good lighting, although as can be seen from the two frames above, it can pull usable detail out of footage that is less than optimum as far as exposure.

Since true RAW data is not output by the RED ONE, the results are subject to changes in the "Builds" of the de-mosaic (de-bayer) software and REDCODE firmware etc. If you have a true RAW camera the data quality is independent of the camera revisions.

In other words, you cannot avoid artifacts of REDCODE and the other software used in the sensor data path at this time. If true RAW data was output then you could write your own software to process it to get the look you want within what data the sensor can produce.

Light interacts with the dyes and colors of the subject and each combination interacts with the dyes in the sensor's filters. Your eyes interact with the type of monitor used, or the dyes in the filmout. Some people see colors not quite the way others see them.

You can Grade the same shot many ways...

Art is not a test chart.

An artiste needs to pick the colors for his palette.

Three strip Technicolor had issues with colors not looking in the print like they did on the set, so they did tests and used colors on the set that would end up the color you wanted in the print later. To some extent that is still true, you need to light, costume, paint, and grade to work to get the colors right in the film out print on a particular film stock, no LUT can make up for all the variables and still give full use of the total gambit the film print stock can display. A LUT can align the center part of a monitor gambit to a print, but that does not deal with the subject's colors and camera artifacts.

Today a DP's job is harder since there are new ways for images to be recorded and exhibited, you do not have time to learn how to control the medium like you could when there was one camera film and one printstock for many years, and you could work with a LAB to get to know how the grading would turn out.

Without true raw data free of build dependent changes, and a stable 3200K tungsten studio lighting, I think you just need to deal with the pixel values in each frame as best you can... you cannot "one light" grade location footage.

My software can batch process frames, and can record the CC from one shot for use with other shots, but there will need to be shot to shot adjustments to optimize the corrections for the subject and lighting interactions... how could it be otherwise?

Camcorders try to normalize footage with "automatic exposure" but that is not Digital Cinema, Digital Cinema is the control of exposure by the DP, full manual exposure, full RAW data, full grading control in post...

Sorry for the long answer, but there are reasons not to look for a push button answer to making a work of art, in addition to the things we cannot change about the early part of the workflow.

 

[David George]

I appreciate your perspective that all shots need grading, but it seems like almost all RED footage has this strange green shift in the skin tone shadows - that seems like a problem that should have some universal solution...

 

[Dan Hudgins]

MAC issues? and Dark Background frame 1478

MAC issues? and Dark Background frame 1478

I appreciate your perspective that all shots need grading, but it seems like almost all RED footage has this strange green shift in the skin tone shadows - that seems like a problem that should have some universal solution...

I have processed Dark Background frame 1478. This image shows the degree of finesse required to get a usable result from some shots. When going to a filmout the dark noise would be diminished since print stock bleeds in the dark areas and would blend the noise together somewhat, so you would NOT want to "crush the blacks" any more since the sholder of the print stock will do that some anyway, and you want to be able to see the outline of the subject on the right's sholder in shadow against the black background (lower right corner of the image).

Your monitor will need to be gamma 2.2 to 2.4 and the brightness set so that you can just see the right subjects sholder in shadow against the black background, or this will look like more of a mess than it needs to... You should also just be able to see the chin shadow of the left subject on his black shirt.

About the greeish grading people end up with. I think something people with MACs are doing to the images may be adding artifacts that are not in the R3D files themselves, or they are not using the neutral settings in REDCINE.

I have been outputting from REDCINE like this, netural settings, REDLOG 16bit TIF, and then grading in my programs. If you have a R3D file that shows greenish artifacts in the images you can post them and I will look and see.

I did have a problem with some footage I shot with my brother on film, but the green shadows were caused by mixed illumination, if you have any florescent lights around giving greenish fill light, and use some other kind of light for the fixtures, you will balance for the bright areas of the faces, and get off color shadow areas. Never shoot with mixed illumination to avoid this problem, there is no simple fix for it, you can introduce cross over by dropping the foot of the green curve but that will give a purple tint to other dark subjects in the shot, so you would not want to do that to everything if it's source is a lighting error.

There may be a mis match in the foot of the red, green, and blue curves, since the compression used seems to compress the green data at 4x the area resolution. When working on this frame I noticed that the compression artifacts are about 4x the area size in the Red and Blue seperations as they are in the Green seperation, I do not have a save for the full screen display of those images, but if you press Control and R G or B in my grading you can look at the Red, Green, and Blue full screen to see what I am talking about.

As far as this frame goes, you can compare this grading with the original image when you play the R3D file. Adding some fill light would have been a big help, as the compression noise comes up suddenly in the darker areas. If the RED ONE put a log curve on the brightness data before compression the shadow noise should be less that what is now added, maybe.

There seems to be an issue in Dark Background frame 1478 with the dual color lights used in the shadow on the right subjects face from his cap, since I was trying to get a nice tan tone on the left subject I did not alter the matrix to bring the color of the two light's colors closer, but whenever you have two color light sources you will get shadows that are off hue.

It helps me may my programs better to see real world footage, so if anyone have frames they would like to see graded this way and that I would enjoy seeing what can be done. Also if anyone has Super16 or 35mm negative frames I can scan that would also help to be sure all the grading options are going to give results people will find useful, maybe...

Another point, LCD monitors seem to "leak" purple in the shadow areas, since their polarizers are not good, if you cross two cheep polarizers you will see dark purple through them. If you calibrate your monitor with a probe, it may read the purple in the shadow areas and make the compensation to add more green to the shadow areas. When you process your frames you may be adding too much green to the shadow areas.

My fix for that kind of error is to use a CRT monitor for grading, adjust the monitor to track well, and not to compensate anything except the film recorder's exposure settings.

[See the notes above in my other posts about adjusting the monitor and using black paper around the click to enlarge image... See also the Light Background frames 0467 and 1206 on the previous page, you need to log in to see the images]