Welcome to our community

Be a part of something great, join today!

  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

Tested Monstro vs. Helium

First - RED color science prior to IPP2, was, no offense to Graeme, less sophisticated than ARRIs. Hardly a shock since ARRI had an entire team that spent years refining the color rendition of the digital files from their film scanners. Additionally, as others have noted, ARRI tweaked their color matrix in specific ranges to rein in certain hue/luma combinations that would rarely be desired. IMO, this is pretty subtle and for typical narrative work 90+% of the tweaks are what you'd want anyway.

Here's a supposition I'd love to try and prove with proper testing: ARRI's color matrix sacrifices some color precision in and near common skin tones to reduce the deviation threshold, RED's color matrix is simply trying to nail scene referred values. Wouldn't such "tweaking" create odd color shits on Alexas? I think it will, but not by much - and if you never saw that scene in real life, how would you know? Skin tone, OTOH, is something humans are evolutionally trained to perceive very precisely (interesting subject, but beyond this thread), so even subtle shifts are easily noticed.

When making the leading lady look her best, or the model for the product look radiantly healthy, etc is your task as a DP - a little bias toward skin tone rendition sounds like a plus.

It's also worth noting that during production, not always but typically, the image on the monitor is king. You can explain grading, gamma curves, rec709 color, ambient light, debayering and a hundred other things but, especially on spots, the WYSIWYG expectation persists.

Cheers - #19

Yes, this is according to my uderstanding as well.

I have not read enough of Graemes white paper to know for sure, but as Im told red IPP2 colors is pure mathematical curves, no color warping as in warping the shape of the color cube. To me that is a more "accurate" way of doing things than creating a 3D lut that push/pull the colors around. As I understand Arri is doing.

What fascinate me the most is how Kodak, Fudji and the others could 1. even get a colored image out of a plastic strip and silver flakes. Then imagine to make all that, mixing chemical liquids etc so it looks better than real.... To me thats quite an accomplishment. doing it digital seams so much more easy, but I guess not.
 
New 3DLut Creator Pro 1.4.6 green skin removal tests. Its not a proper mathematical Color matrix transform, just another method some users may use. Also, because it isn't a proper mathematical Color matrix transform, it will only work for this specific color correction and scene. Both the Redcine-X reference and the 3DLut Creator images used a"5000" Temp and "-16.00" Tint.


I used a different "A/B" skin tone correction this time with a "HSPe" color model the default for 3Dlut Creator.


Screenshot_51.jpg



Also with a smoother curve for the RedWidegamut colors.


Screenshot_52.jpg




Then I exported the combination of the "RedWidegamut" colors, the custom "S-Curve" and the "A/B" skin modifications out as an "IPP2" Lut similar to "Medium Contrast" and a "Very Soft Roll-Off" into Davinci Resolve.



Next I exported a "Medium Contrast" with "Very Soft Roll-Off".Tiff out of Redcine-X Pro to use as a "Reference.


Redcine-X Reference, with skin tone isolated with Vectorscope. Lips were masked so as to not exaggerate red in the vectorscope.


W001_C005_1014_V32.jpg



Screenshot_59.jpg







My Export Lut out of 3DLut Creator Pro 1.4.6, with skin tone isolated with Vectorscope.



test_for_skin_two_2.jpg



Screenshot_57.jpg




with -2 Log Shadow added .


-2_log_shadow_added_1.4.1.jpg




I think I might still need to de-saturate the skin tone just a little bit more however.
 
Last edited:
HA HA ...thanks sreennivasa. But sometimes it takes 3 or 4 tries to get an "acceptable" grade that I can "leave it alone".
 
Hi Rand,

I don't know shit about color mathematics, like you, I just love to color things and was determined, at the time, to get a better color profile for my Helium. But I discovered several limitations in making a 3D lut, at least in resolve. I'm definitely going to try this program to see if I can do better.

I think you did a good job in making the skin look better in your example but now the rest of the scene has a strange reddish cast to it. This is why I qualify all the colors separately in Resolve and play with all the curves and qualification to neutralize all unwanted casts and get as much color seperation and nuance as possible. I still find that Alexa footage gets you to this neutral place with decent skintones right out o f the camera. I sound like a broken record about this but I'm hoping RED is listening and realizes that fixing the color could be a huge boom for sales. They can keep IPP2 for the hardcore but make another profile that's more idiot proof and great for clients.

Been looking at all this night time footage from the Montstro and, yes, its super clean but the color is still so underwhelming in IPP2. I'm a 100% sure this will be the best and easiest RED footage to color and make it look amazing but right outside of the camera its kind of whatevs. Surely its a lot cheaper to hire the best colorist in the world for a few months to come up with a great profile for a RED camera then it is to build a sensor from scratch every year. It's time RED beats the haters at their own game.
 
Anton,

You`re right about the background being more reddish after the grade. However, that is mostly due to the background colors being so close to the female talent's skin tone. Also notice it hardly changed the color chips on the colorchecker or the blacks of the leather in the chair or of the colorchecker. I like you have both "qualified" and "power windowed" a scene to death.


Case in point.

32058983563_0432db47dd_o.jpg
[/url]Untitled000864003jpg by rand thompson, on Flickr[/IMG]


But just imagine doing that for a couple of thousand scenes of a short film or feature. I just want to get atleast a good starting point so that so much qualification and power windowing are not as necessary. A good colorist, which I am not, is priceless and knowing enough to help collabrate with your coloirist to get the best from a scene is also very important.
 
Last edited:
Exactly Rand what I want is how to get the last image you sent: has white balanced colors without a strong cast with good skintones but without using any windows or qualifyers at all.

There's no way to get this perfectly with any camera but the Alexa is the one that gives the easiest starting point. But I'm convinced this can be done better on a color science level. Just look at the Panavision camera: it's a RED sensor but with a much better color starting point.
 
Anton,

Here's another comparison with the change in the "A/B" corrector in 3DLut Creator for skin tones. This time with no corresponding background/ Skin tone conflict. My Lut seems to have lightened certain "Reds" in this scene, I will have to see what caused it later. I understand what you are saying about achieving a desired look in camera instead of having to work for that same look in post. Unfortunately for some red users for now, Alexa and one of the other manufacturers is the only way to get that, barring making the corrections in post as I see it.

Out of Resolve Red "Medium Con- Soft Hi Roll-off"

Out_of_Resolve_Red_IPP2_Mediun_Con-_Soft_Hi-_Roll_1.1.1.jpg



Out of Resolve 3DLut Creator


Out_of_Resolve_3_DLut_Creator_Pro_1.2.1.jpg
 
Rand...this thread also is not a color grading thread.

And what you did there with the image messes up other things in the signal so may feel good in some shots while messing up the other.
 
Please Horjve, are on this again. Yes the problem is with them and apparently also with you, even if I tried to explain it to you many times over.

No need to explain anything.

I'm going to take your name spelling as humour attempt from this point.

So if the resolution does not matter then this camera is a total and complete waste. As thats pretty much all it gives.


It does not matter much for jump from 3K to 4K. If you are typically shooting Alexa and care about that texture.

And no, that is not all it gives. It gives a larger format.


So I dont see how that would magnify the "perspective". Perspective is not anything DOF related,


Larger format > larger lens crop > wider perspective

Wider perspective > longer lens for the same shot > shallower DoF
 
Hrvoje,


When did you become the color grading Police? Anton and I where discussing the green cast problem he was having and as much as you push your Luts anywhere you can, I mean really!
 
Hrvoje,

When did you become the color grading Police? Anton and I where discussing the green cast problem he was having and as much as you push your Luts anywhere you can, I mean really!


...


/shakes head, exhale/
 
rand thompson;1793652 I where discussing the green cast problem he was having and as much as you push your Luts anywhere you can![/QUOTE said:
I think its interesting to see these images color graded so please do keep it up.
/Andreas
 
Hi Rand,

I don't know shit about color mathematics, like you, I just love to color things and was determined, at the time, to get a better color profile for my Helium. But I discovered several limitations in making a 3D lut, at least in resolve. I'm definitely going to try this program to see if I can do better.

I think you did a good job in making the skin look better in your example but now the rest of the scene has a strange reddish cast to it. This is why I qualify all the colors separately in Resolve and play with all the curves and qualification to neutralize all unwanted casts and get as much color seperation and nuance as possible. I still find that Alexa footage gets you to this neutral place with decent skintones right out o f the camera. I sound like a broken record about this but I'm hoping RED is listening and realizes that fixing the color could be a huge boom for sales. They can keep IPP2 for the hardcore but make another profile that's more idiot proof and great for clients.

Been looking at all this night time footage from the Montstro and, yes, its super clean but the color is still so underwhelming in IPP2. I'm a 100% sure this will be the best and easiest RED footage to color and make it look amazing but right outside of the camera its kind of whatevs. Surely its a lot cheaper to hire the best colorist in the world for a few months to come up with a great profile for a RED camera then it is to build a sensor from scratch every year. It's time RED beats the haters at their own game.

How's this:

2bL39k1.jpg


No whitebalancing, no coloring, just graded (contrast&brightness) in NLE. RWG/Log3G10 to 709 done in NLE.



OMVJ0NH.jpg


This is whitebalanced off the chart, in RCXP, then same "treatment" as the one above...
I think IPP2 is more than just for "the hardcore".

@Everybody
Sorry, had to do it. :smile5:
 
No need to explain anything.

I'm going to take your name spelling as humour attempt from this point.




It does not matter much for jump from 3K to 4K. If you are typically shooting Alexa and care about that texture.

And no, that is not all it gives. It gives a larger format.

The quality increase is quite big. Like more than 25% less noise. And even more if finalizing for 4k.

I don't know how many projects shot on alexas this year I pushed out to UHD. I look at pixels up close uncompressed 1:1 or even zoomed in and have done so for the last 27 years. You can repeat your statement above how many times you want. I see the difference, and I see it every day all day. The high noise of the alexas is what is improved with LF and the XL. So what you are saying Hojvje thats simply wrong. 4.5k alexa will for sure create better keys and have way less noise than XT alexas. And yes a lot of people shoot for texture with Arri cams.

I spent half a year working on a quite large commercial where the client simply did not like the noise. No joke, They where thinking of reshooting the whole thing due to it. Sure it could be shot differently. And I´m sure if shot on XL's or LF´s things would look very diffrently at the same exposures just do to the higher number of pixels. And no upsampling but downsampling.

https://vimeo.com/256573224



Larger format > larger lens crop > wider perspective

There is no "wider perspective". There is is "wider field of view". A Perspective is actually a point.

With the same focal length you get a wider field of view with a larger sensor, but who dictates what focal length to use? I don't know if there is any rules that apply when shooting certain shots that only a certain focal length can be used... Another focal length and stop gives the exact same picture.

Wider perspective > longer lens for the same shot > shallower DoF

It gives shallower dof at the same stop. But as you go to a smaller sensor you can also use a smaller image and you gain speed. You actually gain just as much, so no you don't get shallower dof. Larger sensor actually gives you slower lenses at the same fov.
 
The quality increase is quite big. Like more than 25% less noise. And even more if finalizing for 4k.

I don't know how many projects shot on alexas this year I pushed out to UHD. I look at pixels up close uncompressed 1:1 or even zoomed in and have done so for the last 27 years. You can repeat your statement above how many times you want. I see the difference, and I see it every day all day. The high noise of the alexas is what is improved with LF and the XL. So what you are saying Hojvje thats simply wrong. 4.5k alexa will for sure create better keys and have way less noise than XT alexas. And yes a lot of people shoot for texture with Arri cams.

I spent half a year working on a quite large commercial where the client simply did not like the noise. No joke, They where thinking of reshooting the whole thing due to it. Sure it could be shot differently. And I´m sure if shot on XL's or LF´s things would look very diffrently at the same exposures just do to the higher number of pixels. And no upsampling but downsampling.

https://vimeo.com/256573224





There is no "wider perspective". There is is "wider field of view". A Perspective is actually a point.

With the same focal length you get a wider field of view with a larger sensor, but who dictates what focal length to use. I don't know if there is any rules that apply when shooting certain shots that only a certain focal length can be used... Another focal length and stop gives the exact same picture.

Wider perspective > longer lens for the same shot > shallower DoF

It gives shallower dof at the same stop. But as you go to a smaller sensor you can also use a smaller image and you gain speed. You actually gain just as much, so no you don't get shallower dof. Larger sensor actually gives you slower lenses at the same dof.[/QUOTE]

The Coco Chanel Spot looks great, shot with !? Thx in advance ......
 
Another example Hrvoje.

This film also something I finnished for 4k release shot mostly on alexa minis and Xl´s.

The forest scene for example. I can tell you as much as those keys of the treetrunks where no easy task to pull. Flickering colored light into the scene and out of focus swingshift lens shot tree trunks with far from the best lit blue screen behind them... low lit like dusk. Noise big as cornflakes.
Well, higher resolution would have made wonders. Keying and then downscale is the normal procedure. When you go to 4k with alexas you key and then upscale... the noise problems get magnified, makes things way more difficult. Like for these shots quite a lot of money more difficult.

PastedGraphic-1 by Björn Benckert, on Flickr

image by Björn Benckert, on Flickr

And a lot of times as in the setup above (iPhone pic from prelight day), there is plenty of reasons why you can't have great exposure on the green/blue screen etc. To short distance to the back wall, to small of a blue screen, not enough lights. To much bluespil contaminating the scene, practical lights that you want to expose etc, etc. For such events then scaling down from 4.5K to UHD instead of scaling up from 3.2k simply makes the end result way better.

So even if you don't see the difference between 30% more capture K´s, trust me it´s there and it makes tons of different to the end results. Possibly not when just looking at colors but normally there is more done to images than that.


image by Björn Benckert, on Flickr

Or in this setup with the car and the reflective packages on the back of it (also a iPhone snap from prelight)... Low light to get the lights of the car, bluscreen reflecting into the scene... Noise can get quite dramatic when canceling out the bluespill. but with 30% more K´s, less so.


https://vimeo.com/244859097
 
The Coco Chanel Spot looks great, shot with !? Thx in advance ......


Alexa´s, 3 minis and two XL´s as I remember.

More info here: http://www.syndicate.se/kiss-kiss/

If' I a was allowed to show before and after picks I think it would go straight to the Neat video and Crok matchbox reels. Without saying to much I can say, Picture is post filtered quite a bit.
 
Back
Top