ARRI Christmas LUT

[Hrvoje Simic]

List sent.

Hats off Christoph.

 

[Christoph Tilley]

Ok I will try my best. We are going to stream the shootout live at our Facebook page if you guys want to follow along: https://www.facebook.com/mxrproductions/

 

[John Brennick]

Ok I will try my best. We are going to stream the shootout live at our Facebook page if you guys want to follow along: https://www.facebook.com/mxrproductions/

Great work christoph!! That's really cool of you! Thanks for doing this!

 

[Anton Esteban]

Christoph, thank you so much for making this happen! At this point, I'd settle for real daylight, some kind of daylight balanced light, tungsten, and a night shot with mixed lights. The night shot is key for me. What I'm finding from my own tests is that I'm probably going to have to make 2 LUTS to get as close as I can to the Arri Rec 709 look. I'm finding that the LUT I made works great in daylight but still needs work in mixed lighting situations. Also, hopefully one of you is darker skinned then the other. I still want to tweak this for darker skintones.

If you could shoot this with the Standard and Skintone OPLF that would be a huge plus but I'd settle for the Standard. Finally, please shoot RAW if you can. I want to see how the shots change if I alter the different camera's ISO's.

Hats off to you for doing this!

 

[Marc Wielage]

Christoph, thank you so much for making this happen! At this point, I'd settle for real daylight, some kind of daylight balanced light, tungsten, and a night shot with mixed lights. The night shot is key for me. What I'm finding from my own tests is that I'm probably going to have to make 2 LUTS to get as close as I can to the Arri Rec 709 look. I'm finding that the LUT I made works great in daylight but still needs work in mixed lighting situations.

That's what color-correction is for, not a LUT.

 

[Anton Shavlik]

You likely cannot make a LUT to do this. At best you'll have a lut to turn one specific scene shot on one camera into what that specific scene would mostly look like on another.

The red, green and blue pixels on the two sensors likely have different spectral response, so you have to hope that Red's sensor is more discerning, before it's digitized, to all wavelengths of light, as well as have more dynamic range. We know that if something is blown out on 5DmkIII footage but present on red there'll be no way to match them in post. Likewise if Arri's sensor sees differences in dynamic range or color that red's sensor cannot, the LUT cannot make up for it.

You then have to hope that the color processing makes similar decisions. What do you do with out of gamut colors? Likewise, it was well known that the MX would look green under street lamps, while alexa looked warmish. In this instance it would be good to turn green tones warm, but what about a forest scene where an abundance of green tones are present?

Also, LUTs are a surjective function mapping one set onto another, so you always end up with equal or less elements (in this case hues, tones, saturations) after applying them. Meaning, you have to start with more information before the LUT is applied than you want in your target. If you're targeting rec709 from widegamma/log10 you might be OK but if you start by trying to turn one rec709 source into another you're going to lose some subtleties.

I know r3ds are 16bit with ~16 stops of range and with widegamma/log3g10 you have a really big domain to start from but I believe there's going to be situations where one of these problems kills your LUT.

 

[Anton Esteban]

Geez, you try and do something nice and someone on this board's always got to hate. Yes, obviously creating a LUT is not a cure all for everything but if it makes your life a little easier wouldn't you want to be a part of it? When was the last time you threw on the Rec 709 LUT on an ARRI job and called it a day? Obviously it will always need some form of color correction to get the most out of any clip. I've said that the main reason I made this is to use it as a tool to help me protect my highlights when I shoot with the Helium sensor because it shows you that you actually have a lot more detail in the shadows than you realize. I'm about to post another little test here to show how helpful that really is. I shot a little test where I combined the same R3D clip processed at two different ISO's to get the maximum dynamic range I could get out of a given RAW clip (photographers do this all the time). Now the actual combining of the two clips has proven to be quite difficult but maybe some jedi colorist like yourself can chime in on how to do this better. One clip was processed with my LUT at 3200iso and then I combined this with a clip processed at 200iso. Now this is not HDRx this is the same R3D clip processed two ways and joined together in Resolve with some very basic masks. Now what makes my LUT interesting is my LUT is basically one stop brighter than the RED ISO. Right now, the current RED Weapon firmware will not allow you to shoot past 3200iso. With my LUT I was basically shooting at what RED considers 6400iso, something you can't even do on the camera (I guess you could theoretically use the FLUT control to go beyond 3200iso). I basically exposed using the GIOscope and made sure not to let anything in the shot go deep red so as to not clip and then I let everything else fall where it did. I wanted to see how much I could recover from the shadows. We all know how hard it is to correct for something over exposed but under exposed can often be salvaged with a little creative noise reduction. I didn't use any noise reduction on the clip I'm about to post but I'm working on another shot that basically had me expose the shadows to 6400 (I had to use some noise reduction for this) and I exposed the highlights at 50 iso. How Im combining these clips is still super beta, I only tried this today but I'm sure if more amazing colorists like yourself chime in, we can all figure out a way to get the most out of this sensor.

A note for the colorists out there, I did try recovering the highlights by using the highlight control, soft curve control, and log highlight controls but it wouldn't bring back enough of the highlights. I knew they were in there because I tried lowering the ISO on the clip and, lo and behold, there was plenty of detail there. So combining the two differently exposed clips was my next option. If anyone has an easier idea on how to do this please chime in. Rendering two 8K clips at the same time, especially if there's any noise reduction, has caused my maxed out mac pro 2013 to crash multiple times. I think its time to move to a PC with this 8k business but that's a whole other post...

PS I had my wife wave her hand in front of the window because I wanted to see if I could still bring back the highlights when something was in front of the place I was going to use some HDR masks. It still needs work but I think the results are usable.

https://www.dropbox.com/sh/pf4ife3q2fqc6gv/AAA50zKWfZrfT5DqF-LF01jPa?dl=0

There's 4 clips in this link so you can better understand my process. All the clips are not up yet, my upload speed where I am right now is terrible. All the clips should be up in about an hour.

 

[Anton Esteban]

Also, this LUT is not about making a RED camera into an Alexa. All I wanted to do was make a tool that helped me get more color contrast, more subtle skintones, nuanced tonality, and better mixed light performance. Why did I use an ALEXA as my guidepost? Because its the undisputed king of all those details I just mention. The proof is in the movies out there. How many films are shot on the Alexa vs the RED? RED had its moment with 3D but that fad is long gone so they're getting back in the game with a full frame Dragon sensor and this new low-light 8k Helium monster. All I wanted to do was try and make it better instead of saying that this camera is better than that camera because of this and that. If anyone can post a side by side of the IPP2 vs my LUT where the IPP2 looks better that will help me tweak it to improve it. This is about all of us trying to improve the tools we have. I've been correcting RED footage for over 5 years and I noticed that correcting the footage with charts I invariably kept making the same similar corrections. So I applied that methodology with an eye for how the Alexa handles contrast in different areas of the image. If I can make a LUT in two days that makes an improvement on Helium footage with 2 side by side ALEXA and RED HELIUM shots then what if we all put our heads together and actually pushed to see how far we can make these cameras better. It's time we all took a crack at figuring out ways to help each other get the best out of these cameras.

 

[Hrvoje Simic]

At this point, I'd settle for real daylight, some kind of daylight balanced light, tungsten, and a night shot with mixed lights. The night shot is key for me.

Mixed lighting is a completely unknown basis with most variables and zero reference.

Geez, you try and do something nice and someone on this board's always got to hate.

There is no hate. We are all on the same side.
You are just oversimplifying matters which may easily lead to incorrect conclusions.

A note for the colorists out there, I did try recovering the highlights by using the highlight control, soft curve control, and log highlight controls but it wouldn't bring back enough of the highlights.

I knew they were in there because I tried lowering the ISO on the clip and, lo and behold, there was plenty of detail there. So combining the two differently exposed clips was my next option.

That is a wrong option. And completely unnecessary.
Clips are exposed the same, you are "developing" them with a different mid point.

If anyone has an easier idea on how to do this please chime in.

Things of value are seldom made by taking the easier route.
The idea alone wouldn't help you much. The solution might.

 

[Anton Esteban]

I explained that I did try many different options to recover the detail and they didn't work, this is the only solution I was able to come up with. One thing that i discovered that was really interesting is I ended up realizing that you can bring back more highlight information,

even below ISO 50 if you convert the gamma to HDRST.84. The problrm is the shadows are completely crushed and thus the best solution I could come up with was mixing the two exposures. Now simply editing this in an HDR standard would solve this problem but none of my clients are clamorng for this so I'm trying to see if there's a way to squeeze as much dynamic range into Rec 709. I'll post the R3D file so you or any other colorist out there can show us how to do it better.

From here on out can I just ask that people actually try and show how to do this better than just say it can't be done. Show us what you got red users! Post me examples where my LUT fails, show me a better way to recover highlights. Put your money where your mouth is!

 

[Anton Esteban]

Ok here is another example of my HDR test. In this clip, I exposed my shadows at 6400 ISO and the highlights at 50 ISO (using HDRST.84). I'd love to post some Pro Res files along with the R3D's but my internet right now is so slow so that's going to have to come at a future date.

Again, there's no substitution for properly lighting a scene to get the best exposure for any given scenario but we've all been in that spot where we just can't do that. I've seen so many RED films where I notice the highlights blowing out in situations where I knew the Alexa wouldn't. This test shows you that you can get a lot more highlight retention from the helium sensor by exposing to the left. I'm using these tests to see where that limit is. This shot is, in my opinion, is pretty much where you could push it. If this was a night scene it would be too noisy, even with really good noise reduction. But, in a daylight scene, I think pushing the ISO this far can still be salvaged. I essentially pushed the ISO by 8 stops. That's a pretty impressive dynamic range. Now, getting it into one R3D at one given ISO and then correcting it in the Rec 709 space is where I'm having problems. Exposing the R3D twice and then combining it in a program like Da Vinci Resolve is far from ideal.

https://www.dropbox.com/sh/0ygzma7dhg781xy/AADRfKhQt5yQuiWrf_EO52rza?dl=0

 

[M Most]

I explained that I did try many different options to recover the detail and they didn't work, this is the only solution I was able to come up with. One thing that i discovered that was really interesting is I ended up realizing that you can bring back more highlight information.....
From here on out can I just ask that people actually try and show how to do this better than just say it can't be done. Show us what you got red users! Post me examples where my LUT fails, show me a better way to recover highlights. Put your money where your mouth is!

Allow me to quote myself, via an excerpt from a blog post I wrote way back in 2011:

The LUT approach is often misunderstood by those who feel, for some reason, that what a LUT is supposed to do is yield a “perfect” starting point for every image. That is not the case. What a LUT does is transform the log curve back to a proper gamma corrected curve based on the particular display device being used, most often a monitor. In the case of an image that is a bit underexposed, this will likely cause the blacks to be crushed by the LUT, and in the case of an overexposed image, it will cause the whites to be clipped. This is where the misunderstanding comes into play. The LUT is not meant to be used as the first correction in a grading pipeline. If you place the LUT after an initial correction, you can then raise the blacks or lower the whites on the original log image prior to it being transformed by the LUT. The LUT will still do what it’s supposed to do, and you then have the option of grading prior to the LUT or after it as needed. This is the “secret” that many professional colorists (especially those who work in the digital intermediate world) know, and few inexperienced ones seem to understand. A LUT is not a color correction, just as a log image is not an image. It is a transform, and the log image is a container. The image needs to be “processed” by the LUT in order to appear properly to the eye.

In other words, an exposure correction is often necessary to put the values of the incoming image into the range expected by the LUT, which has hard coded, predefined values.

Does that help?

 

[Anton Esteban]

Exactly!!!!! Here's a trick with any LUT, not just the one I posted. Put a node before the LUT and use the shadow/Highlight control in Da Vinci resolve and see how much more dynamic range you can squeeze out of any given LOG encoded clip, You'll be amazed at how much you can salvage. And this trick works for any camera as long as the clip is encoded in some form of log or linear format.

 

[M Most]

Exactly!!!!! Here's a trick with any LUT, not just the one I posted. Put a node before the LUT and use the shadow/Highlight control in Da Vinci resolve and see how much more dynamic range you can squeeze out of any given LOG encoded clip, You'll be amazed at how much you can salvage. And this trick works for any camera as long as the clip is encoded in some form of log or linear format.

It's not a "trick." It's an implementation of a proper color pipeline. It's hardly specific to DaVinci Resolve or any other grading program. It's exactly what a "proper" DIT should be doing on set as well, applying a CDL based correction prior to a "normalizing" LUT when working with a camera's log output.

A more general statement would be that if there is any "trick," it's that people who want use to LUTs should educate themselves as to exactly what a LUT is and what it's meant to do. And people who are performing color grading tasks should educate themselves about what a proper color pipeline is. This statement is not personally directed at you or anyone else here, it is a general statement about people making assumptions about technical things when they should be trying to understand them first.

 

[Anton Shavlik]

Anton, you should try starting with log3g10. It would be really rare to need to mix two ISOs in order to capture all the highlights before clipping with this log format. Also if you start with RedWideGamutRGB you'll have the most colors available to pull from.

I don't think people are saying you shouldn't make a LUT that's to your taste, or your client's taste, on highlights, skin tones, curves and so on, to be used to monitor or be implemented in a color pipeline. Making one for daylight, one for tungsten, and one for night or lowlight scenes is smart as well.

It's just not going to make a Red match an Arri, anymore than a lut can make a 5DmkIII match Weapon VV, or even vice versa.

ACES is actually another route to consider, since it attempts to reverse a lot of decisions made by the camera manufacturer and by the sensors. Of course it makes its own decisions instead.

 

[Anton Esteban]

Thank you for clarifying this. I'd love to hear what are the best practices for using a LUT in a color correction pipeline. This would help a lot of users on this board. Normalizing LUTS are the cornerstone for many camera platforms. I'm sure there's plenty of confusion out there on how best to use them. In my case, I've learned a lot about how it works from mixinglight.com, the color correction board liftgammagain, creative cow, and the arri website. Below is a list of videos that helped me understand how to best work with LUT's. Unfortunately mixinglight.com is a service I subscribe to so you have to pay to watch them. But I'm sure there's plenty of other free resources out there that you will so kindly share with all of us so we can better educate ourselves about the best way to implement LUTS in a color correction pipeline.

https://mixinglight.com/portfolio/4-things-to-know-about-working-with-luts/

https://mixinglight.com/portfolio/difference-between-1d-and-3d-luts/

https://mixinglight.com/portfolio/matching-shots-with-luts/

https://mixinglight.com/portfolio/using-arri-alexa-luts-part-1/

https://mixinglight.com/portfolio/lut-preppers/

https://mixinglight.com/portfolio/reverse-engineering-luts/

https://mixinglight.com/portfolio/rolling-your-own-luts/

https://mixinglight.com/portfolio/working-with-luts-a-3-node-approach/

 

[Anton Esteban]

Anton, I always use log3g10 and RedWidGamutRGB for all my RED corrections. RED's implementation of this is a revolution, in my opinion, for anything ever shot on RED. That being said what other tools can you use to compress the dynamic range as much as possible into on shot in REC 709? I find pushing the file in either direction either makes the shadows too dark or the highlights too bright. I've tried playing with curves, LOG controls, primary controls, contrast controls, shadow and highlight controls, qualifying ranges of lumanosity and they all didn't really get the job done. I know that the ISO controls just moves the midpoint along the log file or just adds a certain amount of gain to a given LOG range but, invariably, I can never seem to get all the information in one set ISO point. I know the real solution here is to correct in an HDR space. The new Helium sensor has a lot of dynamic range that can fit into the HDR sphere. I just wish there was a way to make it work in a REC 709 standard which is still, unfortunately, the norm for most projects.

 

[M Most]

Thank you for clarifying this. I'd love to hear what are the best practices for using a LUT in a color correction pipeline. This would help a lot of users on this board. Normalizing LUTS are the cornerstone for many camera platforms. I'm sure there's plenty of confusion out there on how best to use them. In my case, I've learned a lot about how it works from mixinglight.com, the color correction board liftgammagain, creative cow, and the arri website. Below is a list of videos that helped me understand how to best work with LUT's. Unfortunately mixinglight.com is a service I subscribe to so you have to pay to watch them. But I'm sure there's plenty of other free resources out there that you will so kindly share with all of us so we can better educate ourselves about the best way to implement LUTS in a color correction pipeline.

My blog (rarely updated, but still valid) is mikemost.com. Everything I've mentioned here and more is discussed there.

 

[Anton Esteban]

Wow Mark, this is really helpful, especially the article about 'understanding log grading'. Required reading!

 

[Hrvoje Simic]

Exactly!!!!! Here's a trick with any LUT, not just the one I posted. Put a node before the LUT and use the shadow/Highlight control in Da Vinci resolve and see how much more dynamic range you can squeeze out of any given LOG encoded clip

If that LUT transforms any particular (non-global) color values and is not done properly, any exposure compensation prior to the LUT will have differing and unexpected color results after the LUT. So it's not that simple.