ARRI Christmas LUT

[Christoph Tilley]

Ok, we did this! We shot different cameras in the studio:
- ARRI Alexa Mini (ARRIRAW, Open gate 3.8K)
- Sony F65 (F65RAW-SQ)
- RED Epic Dragon (6K, 5:1)
- RED Weapon Helium (8K, 5:1)


Everything was shot on a ZEISS LWZ.2 at 45mm. We measured T8 at our models face. All cameras were shot at following T-stops: 8, 11, 16, 22, 5.6, 4, 2.8
Every camera was set to ISO800, 5600K, 25FPS and 180-degree shutter.

In Davinci Resolve we applied following LUTs to the raw image data:
- Alexa LogC -> Arri Alexa LogC to Rec709.dat
- F65 SLog3/SGamut3.cine -> SLog3SGamut3.CineToLC-709TypeA.cube
- Dragon RedLogFilm/DragonColor2 -> RLF_2_RG4_64.cube
- Helium REDWideGamutRGB/Log3G10 -> IPP2 Preview.cube
- Helium REDWideGamutRGB/Log3G10 -> Antons_ARRI_Christmas_LUT_BETA_v1.cube


We are sharing in our dropbox:
- the resulting clip where everything is thrown together
- complete raw footage
- uncompressed dpx still frames straight out of resolve
- jpg still frames for reference
- all LUTs used in resolve

You can download everything here:
https://www.dropbox.com/sh/1m8md5zirddrt2t/AACQqFhKPZN7WYtW0Gprzq3da?dl=0

If you like this, please consider giving us a like on our Facebook page:
https://www.facebook.com/mxrproductions/

Happy holidays everyone!

 

[Christoph Tilley]

Quick conclusion from my side: I REALLY LIKE what Anton did with his ARRI LUT. It gives you one more stop light digging out those clean Helium shadows. I will definitely shoot with this LUT on the next gig. Only if someone could fix the green cast on the REDs :-(

 

[Nate Maydole]

Looking forward to giving this a test with my Epic-W. Thanks for posting.

 

[Anton Esteban]

Christoph!!!!! Wow, this is such a holiday treat!!!!! I'm going to like that Facebook page to no end!

 

[Anton Esteban]

On a side note, here is a link to a shortened R3D file of that clip I was trying to bring back the highlights for anyone out there who can show me a better way of compressing the helium sensor's large dynamic range in a REC 709 space.

https://www.dropbox.com/s/ewxevb948op8yec/A001_C011_1227DQ_001.R3D?dl=0

 

[Josiah Duncan]

Really digging Anton's lut on some of my already shot Epic-W footage i've been casually playing with the past few days. IPP2 is great in it's own way so I could definitely see myself using both depending on projects.

 

[rand thompson]

On a side note, here is a link to a shortened R3D file of that clip I was trying to bring back the highlights for anyone out there who can show me a better way of compressing the helium sensor's large dynamic range in a REC 709 space.

https://www.dropbox.com/s/ewxevb948op8yec/A001_C011_1227DQ_001.R3D?dl=0

Anton,

here is the original file from your dropbox and Parade scope.

[/url]Untitled00086411jj by rand thompson, on Flickr[/IMG]

[/url]starting point by rand thompson, on Flickr[/IMG]



here is the original file from your dropbox plus your Arri lut and Parade scope.

[/url]Untitled00086400jj by rand thompson, on Flickr[/IMG]

[/url]arri lut by rand thompson, on Flickr[/IMG]

here is the original file from your dropbox and a manual grade with the usual suspects (qualifiers and power windows)and Parade scope.

[/url]Untitled00086404jj by rand thompson, on Flickr[/IMG]

[/url]manual grade by rand thompson, on Flickr[/IMG]

F****d up qualifying the wall from green due to the RGB noise from raising the shadows. Also, using free vesion of Resolve so no noise reduction.

 

[Anton Esteban]

Helium camera owners, I need to ask a favor. Can you post some R3Ds of footage you've shot in different lighting situations. I want to improve the LUT and test it on as many scenarios as possible. Especially post clips that your having problems with or are not happy with the color out of the box. When I made my first LUT I tested it on as many R3Ds I could find on reduser but they were all shot on the old firmware. This time I want to test it with the new firmware.

 

[Anton Esteban]

This looks good Rand. I think I need to rethink my methodology. One thing I realized from your example is what if you added some contrast to the power windows you used on the windows and you played with the contrast pivot to get it where you think it looks good. Maybe that would bring back even more detail and color form the brightest parts of the image. Have you tried bringing the clip down to ISO 50 and then using HDRST.84 to the gamma? Look at how much info and detail there is in the windows. That's what I'm trying to bring back nto the image. Also, have you tried using Neat Video Noise reduction? I think it should work in Da Vinci Resolve lite. It takes forever to load but if you mess around in the advanced section of the noise control I've gotten the best results out of any other noise reduction tool I've used. You can even refine it even more by qualifying only the shadows, midtones, or highlights depending on where the noise is worst. I recommend using the 'keep more detail' setting in Neat Video if you ever decide to get the plugin. It looks the most 'natural' to me.

 

[Anton Esteban]

One more question for you Rand, did you apply my LUT to the clip using RWG and Log3G10? It looks very contrasty and over saturated. Definitely not what I got when I used it on my end.

 

[rand thompson]

Anton,

You're right, used the wrong file.....REDO!!!



RedwidegamutRGB-Log3g10 + Anton's ARRI Lut and Parade Scope.

[/url]A001_C011_1227DQ by rand thompson, on Flickr[/IMG]

[/url]arri lut redcine by rand thompson, on Flickr[/IMG]

I tried most of what you suggested.


Here is brightness inside of the Power window lowered , Notice the now black window frames with little change in overall contrast in the background.

[/url]Untitled00086401jjj by rand thompson, on Flickr[/IMG]


Next tried a Luminance and HSL Qualifiers to only isolate the bright background inside the Power window leaving out the Window frame and Star.

[/url]power window lum qual by rand thompson, on Flickr[/IMG]



Here's what I got after doing so.

[/url]Untitled00086405jjj by rand thompson, on Flickr[/IMG]


And contrast and pivot just blew out the highlights all over again.


I will be looking into neatvideo or most likely the Studio Version of Resolve


What I did choose was this basic curve inside the Power windows.

[/url]my curve by rand thompson, on Flickr[/IMG]

[/url]Untitled00086400jjjj by rand thompson, on Flickr[/IMG]

 

[Dana Rice]

This is my first time working with LUT's, so I have a few questions. My normal workflow is to start with a RG4/DC2 for shots without a lot of contrast... for ones where I need to pull more detail out of the highlights/shadows I use RWG/LG310 and simply grade with those as my starting points. The last few days I've been playing around with Red's IPP2 LUT as well as Anton's. In all of my tests so far, I find that I do prefer Anton's LUT to Red's in almost every case... however, I also find that any shot's where I need to get detail out of the shadows/highlights, I am simply better off not using either LUT and just grading from RWG/LG310 with no LUT as I have been before. As I can't seem to pull as much detail and get as good of a grade with the LUT applied. Also, with shots that don't have a bunch of contrast in them, I still find DC2/RG4 to be the best starting point. So I guess I'm just wondering what the point of LUT's are beyond giving you a non-flat preview image for yourself/clients while shooting, which DC2/RG4 already seemed to do just fine at anyways. I'm wondering if I'm missing something or doing something wrong in my grading that's limiting the usefulness of these LUT's.

 

[Hrvoje Simic]

So I guess I'm just wondering what the point of LUT's are beyond giving you a non-flat preview image for yourself/clients while shooting, which DC2/RG4 already seemed to do just fine at anyways.

The point of a LUT is having a very precise and particulars-oriented, multi-platform multi-tool applicable image/signal transformation for a specific purpose, base and outcome.

I'm wondering if I'm missing something or doing something wrong in my grading that's limiting the usefulness of these LUT's.

You are asking a specific response on a very general question.



Anton, the reason you are having issues with correcting that shot is because it is technically incorrect and underexposed 3-4 stops and pulling those shadows up is just the abuse of 16 bit sampling.

 

[rand thompson]

This is my first time working with LUT's, so I have a few questions. My normal workflow is to start with a RG4/DC2 for shots without a lot of contrast... for ones where I need to pull more detail out of the highlights/shadows I use RWG/LG310 and simply grade with those as my starting points. The last few days I've been playing around with Red's IPP2 LUT as well as Anton's. In all of my tests so far, I find that I do prefer Anton's LUT to Red's in almost every case... however, I also find that any shot's where I need to get detail out of the shadows/highlights, I am simply better off not using either LUT and just grading from RWG/LG310 with no LUT as I have been before. As I can't seem to pull as much detail and get as good of a grade with the LUT applied. Also, with shots that don't have a bunch of contrast in them, I still find DC2/RG4 to be the best starting point. So I guess I'm just wondering what the point of LUT's are beyond giving you a non-flat preview image for yourself/clients while shooting, which DC2/RG4 already seemed to do just fine at anyways. I'm wondering if I'm missing something or doing something wrong in my grading that's limiting the usefulness of these LUT's.

Dana,

Just like you I didn't use Luts and just started from a flat image such as V-log and S-log. From my limited experience with them( didn't know what they were and how to use them until about a year and half ago) I believe to get more out of them you have to do a "pregrade" before you apply the lut. For example, make sure you have your initial image properly exposed and white balanced(RGB channels normalized)...etc. Next, apply the Lut and use one node away from the initial image and before the addition of the Lut so that you can counteract anything you don't like about the Lut.

Luts are basically "set in stone" predecisions about how an image should appear to look. If you apply a Lut to the initial flat image, any further grading would basically be "deconstructing" the image instead of "building up" the image. DragonColor 2 - Redgamma 4 is like a pre-built High-performance vehicle already designed to meet automotive standards and specifications. RWG/LG3g10 is like all of the parts of the car, but you decide "gear compression ratio", engine-displacement, "power to weight ratio, "timing", Suspension setup and so forth and so on. A lut is deciding if you want a 4-cylinder, 6-cylinder or 8-cylinder engine, that you can modify within certain limitations.

There are far more qualified Red users with far more experience with Luts than myself and hopefully they will chime in to give you a much better explanation.


Rand

 

[Hrvoje Simic]

If you apply a Lut to the initial flat image, any further grading would basically be "deconstructing" the image instead of "building up" the image.

That is incorrect. Whether that would or would not be the case, would depend on what the particular LUT actually does.

 

[rand thompson]

Hrvoje,

Meant to clarify a "Creative/Artistic" Lut and not so much as a "Technical/Transformative" Lut.

 

[Anton Shavlik]

LUTs can do lots of things, they're not specifically for targeting one look or display color space.

LUT stands for Look Up Table. It's two lists of numbers, and it says to change the first number of list one into the first number of list two. "If you see the number 1, change it to 3. If you see the number 4, change it to 7. If you see the number 8, also change it to 7"

These video LUTs are specifically looking for RGB numbers. 3D LUTs turn one set of 3 numbers into another. A LUT that lifts your blacks will say "If you see 0,0,0 for RGB, turn it into 10,10,10. If you see 1,1,1 also turn it into 10,10,10" etc. It just goes through each pixel in your image and checks its lists to see what to change it into.

You can make a LUT to invert the image "Turn 255,255,255 into 0,0,0 and 254,254,254 into 1,1,1" etc, or to make only the blue channel appear "turn 1,1,1 into 0,0,1 and turn 255, 32, 55 into 0,0,55"

To me thinking about the fact that I'm just making these types of lists helps me know when a LUT is useful or adequate or whether more complicated decisions need to be made. Like with Anton's shot of the windows and table, the LUT can't know if it's a dark tree trunk that's over exposed or a white reflective object exposed normally, so any attempt to make a LUT that brings back over exposed windows will probably not work well with objects that are naturally bright, etc.

I'm sorry I know this is common knowledge to most people here but for those reading along this is why a LUT alone likely cannot turn a shot from one camera into another. There are decisions made before those RGB numbers are calculated that a LUT cannot undo. Just like if I were to ask for the square root of 1, there are two correct answers, 1 and -1. A LUT cannot make contextual decisions for where 1 or -1 makes more sense, it has to pick one or the other and stick with it for every pixel.

 

[Anton Esteban]

The example I posted, that Rand was so kind enough to take a stab at, was an extreme test to see if you could salvage a shot that is underexposed by 3-4 stops and the answer is yes. It's not pretty but it can be done. Based on seeing how Rand took a stab at correcting the R3D, I found other ways to bring back detail and color into the overexposed windows. I Found that by playing with the saturation vs saturation curves, and using the contrast and pivot tools within a node that a power window with a heavy feather, and using some good old fashioned lift, gamma, and gain you can bring a good amount of highlight detail back into what seemed like an impossible shot to use. This was a test to figure out the limits of the Helium sensor and, judging by this small test, 3-4 stops seems like the limit.

Now, the LUT I made helped me see in camera that there was still enough detail in the shadows to even attempt to use it. Shooting that many stops underexposed is not a smart move but there are definitely times where you just need to find a way to somehow keep those highlights in a shot and at least you now know that you could do it. Not ideal but it can be done.

 

[rand thompson]

The example I posted, that Rand was so kind enough to take a stab at, was an extreme test to see if you could salvage a shot that is underexposed by 3-4 stops and the answer is yes. It's not pretty but it can be done. Based on seeing how Rand took a stab at correcting the R3D, I found other ways to bring back detail and color into the overexposed windows. I Found that by playing with the saturation vs saturation curves, and using the contrast and pivot tools within a node that a power window with a heavy feather, and using some good old fashioned lift, gamma, and gain you can bring a good amount of highlight detail back into what seemed like an impossible shot to use. This was a test to figure out the limits of the Helium sensor and, judging by this small test, 3-4 stops seems like the limit.

Now, the LUT I made helped me see in camera that there was still enough detail in the shadows to even attempt to use it. Shooting that many stops underexposed is not a smart move but there are definitely times where you just need to find a way to somehow keep those highlights in a shot and at least you now know that you could do it. Not ideal but it can be done.

Anton,

Glad I could be of some help.

 

[Marc Wielage]

Just like you I didn't use Luts and just started from a flat image such as V-log and S-log. From my limited experience with them( didn't know what they were and how to use them until about a year and half ago) I believe to get more out of them you have to do a "pregrade" before you apply the lut.

One interesting thing is that if you "pregrade" enough, you'll find you've corrected the scene without a LUT. Try two nodes in a row and match the gamma and colorimetry that way. It can be done most of the time, particularly for just a straight camera image without a heavy stylized look.