RED Color - A look at REDWideGamutRGB/Log3G10

[Antony Newman]

(Hunch) Another random thought that might illustrate colour differences.

Do we need an image with the colour checker at different angles of incidence to the camera?

Directly facing, 30deg, 60deg.

AJ

 

[Dominik Bauch]

Awesome.. exactly why these discussions are so important. I am just so tired hearing how Arri has such a better look, Years and years of hearing it now... And the sucky part is for the most part they are right.. I think sometimes we get so lost in the pixel peeping that we forget that we are artists first. An image may not 100% reflect what exists, but thats what cinema is all about... creating beautiful images, pleasing to the eyes... My personal hope is after you are done with your magic making, the Arri color bandwagon can finally come to an end... Anyways thanks for everything you are doing!

Totally agree. Nicely put.

 

[paulherrin]

quick grade w/ gamma pushed towards red/magenta and a little saturation offset towards blue
this is where it looks more natural to me while maintaining your look

 

[rand thompson]

here is another grade, however I think it might show up a little pink on some of the calibrated monitors of some members.

[/URL][/IMG]

 

[Hrvoje Simic]

Guys...few things useful to note.


1) The color tonality in Phil's shot of brunette girl greatly depends on artificial light spectral response + ambient light bias.


2) Defining image properties based on that shot gives insight of that shot and that light, not the camera. Grading it based on that shot gives result which works for that shot and those light properties. Not generally.

Camera response is quite a bit more complex matter.


3) Expecting the camera or the transform to provide perfect results in every light scenario is an unrealistic expectation. This is the job for the colorist, to compensate to a possible extent what specific light values offset.


4) What you see is a combination of:

a) What your vision is influenced by
+
b) What your brain expects to see
+
c) What is actually there



5) How you valorize what you see is based upon:

a) Sensibility
+
b) Taste, criteria
+
c) Reference



6) How you desire that skin to look under that light and how it actually does look are two different things.

Former cannot be the reference.



7) If that chart patch is white, that top isn't.

One test transform example.



6) Toning the image for that top, or guessed/desired skin tone of that shot is a job for the colorist. Different priorities, tools, methodology and results.

 

[rand thompson]

Guys...few things useful to note.


1) The color tonality in Phil's shot of brunette girl greatly depends on artificial light spectral response + ambient light bias.


2) Defining image properties based on that shot gives you image properties of that shot, not the camera. Grading it affects image properties for that shot.

Camera response is quite a bit more complex matter.


3) Expecting the camera or the transform to provide perfect results in every light scenario is an unrealistic expectation.


4) What you see is a combination of:

a) What your vision is influenced by
+
b) What your brain expects to see
+
c) What is actually there



5) How you valorize what you see is based upon:

a) Sensibility
+
b) Taste, criteria
+
c) Reference



6) How you desire that skin to look under that light and how it actually does look are two different things.

Former cannot be the reference.



7) If that chart patch is white, that top isn't.

One test transform example.



6) Toning the image for the shirt, or guessed/desired skin tone for that shot is a job for the colorist.
Different priorities, tools, methodology and results.

Here was my logic for the grade you see above.

from looking at the the image , I determined the it was divided into two separate lighting planes. The first lighting plane was the key light that illuminated the left side of her face and the datacolor spyderchecker. the second lighting plane was from behind her head and behind the datacolor spyderchecker.


I felt in order to correct the entire image from the spydercube and spyderchecker, that all of the light in the image had to be illuminated on both the spydercube and spyderchecker. From my comments above, I felt that was not the case. So to color correct the image based on only the Key light which illuminated the left side of her face and the datacolor spyderchecker and spydercube and not what I believe was the tungsten light from vanity mirror in the background that was not captured by the spyderchecker or the spydeycube, would result in a less than optimized color corrected image.


I believed that by using only the datacolor spydechecker and sypercube from one of the two lighting planes would result in an overall incorrect color grade for both of the lighting planes.


Here is the grade including the datacolor spyder checker and spydercube along with and added contrasty datacolor spyderchecker.

[/URL][/IMG]

[/URL][/IMG]

[/URL][/IMG]


****CAVEAT- I used an uncalibrated laptop screen to make this color correction so the colors may be okay or completely f**kup*****

 

[paulherrin]

didn't mean to turn this into a gradeMe thread, my bad... my current understanding of the red wide gamut is that it's supposed to hold for display a wide range of color & light.
anybody can desaturate an image and hide a lot of color casts or grading mistakes especially if it's just a still.
besides improved highlight rolloff, what i'm seeing on the vectorscope is a 'thick' range especially noticeable in skintones, but what i don't find appealing is where it goes 'below' the orange/skintone line
maybe it should 'rest' on the line instead of being evenly distributed above/below

 

[Phil Holland]

I'll give you a hint on that scene since I shot it. You should be using the patches for tonal reference and that's about it. Try white balancing to her shirt. There's more going on in that scene than meets the eye lighting-wise. Yes I know I've been bad.

 

[Phil Holland]

Graeme has been hard at work btw.

 

[Phil Holland]

Might be time to post a few things. Give me a minute.

 

[rand thompson]

I'll give you a hint on that scene since I shot it. You should be using the patches for tonal reference and that's about it. Try white balancing to her shirt. There's more going on in that scene than meets the eye lighting-wise. Yes I know I've been bad.

I didn't mean to turn it into a "Grade-Off" either, I apologize If I did so.

Phil, I will check that out, but promise you I won't post the grade. Thank you.

 

[Mike P.]

Might be time to post a few things. Give me a minute.

Hitting refresh repeatedly...

 

[Phil Holland]

Time for another image drop.

Graeme's been cranking away hard and doing great things.

I've been testing out about 300 images plus the footage I'm currently shooting with the REDWideGamutRGB/Log3G10/Transform LUT method. A large variety of subject matter, light sources, ISOs, etc....

The short things I can say at the moment is that Graeme is after:
- Better Color
- Better Skintones (more subtle hue variations as well)
- Much better handing of out of gamut colors

This has been a process of revision, adjustment, and refinement thus far, but he's onto something great. He's been working his butt off and there's more to this method than even meets the eye.

For the time being, here's 20 samples to look at, some have had an additional curve thrown in as I'm testing gradability as well.

I think Graeme would be curious to know which letter you prefer if you have a moment. I know which one(s) I like

1.

2.

3.

4.

5.

6.

7.

8.

9.

10.

11.

12.

13.

14.

15.

16.

17.

18.

19.

20.

 

[Visar Sulejmani]

I personally prefer D & B.

 

[David Grigorian]

A C E F G and sometimes H

 

[Nathan DuMoulin]

I personally prefer D & B.

B & D are actually my least preferred in terms of skin tones. Particularly B on images 6 and most notably 9.

Speaking of which, it would be helpful if you could number the images Phil.

Either way, thanks for your contribution and hard work (applies to both Graeme and Phil).

 

[Hrvoje Simic]

Are RWG & Log3G10 defined and locked and can be counted on as a basis or they are still worked on ?

 

[Phil Holland]

*added numbers to below the images.

 

[Akin A]

At first glance, I prefer A, C, H but I would probably inspect more before deciding for sure.

 

[David Grigorian]

If I had to pick an overall favorite for all of the images it would be a tossup between E and G. the only ones I don't like are B and D specially on skin tones. To be honest though I wouldn't mind having all of them as an option to play with.

Thank you Phil and Graeme for all your hard work, really appreciate it.