Your LUT creation process (6 replies and 5 comments)
I've been wanting to ask you these questions about your process of LUT creation for a while, it would fantastic if you can shed your light on these =) Big thanks in advance!
1) What is your process of creating your LUT. I understand that you mentioned that you give the image of a little contrast and warm up the shadow and the desaturate a bit, but I guess what I'm ask would be more along the line of the actual process/workflow. (e.g. light a test scene with a certain look and record - then take the Log C/Raw footage to convert to REC709 and add your changes upon that - then load into display monitor onset for reference?)
2) What program do you use with your DIT to create these LUTs?
3) Do you eventually grade your footage exactly like your LUT or do you just use your LUT as a very close reference as to what you would like to do with that RAW footage of yours? and maybe do minor changes and tweaks in post to your likings?
Apologize for bombarding your with three questions at once. I'd love to hear your thoughts!
Hi Yao, this thread has a lot of information about LUT creation:
From reading through those threads on LUTS i have a question i would love someone to answer. I would like to create a LUT to use in conjunction with Log C on the Alexa. I would use the LUT on a 3D LUT capable monitor rather than in the camera.
As i understand it, there are two basic ways you could create this LUT.
1. Creating a LUT by taking an Arri LUT/3D LUT from their website and changing the values in code.
2. Creating a look/grade in Da Vinci Resolve, saving it as a 3D LUT and then importing that into a 3D LUT capable monitor or Arri Amira via the Arri Color Tool.
My question is how viable is approach number 2? Using Da Vinci resolve to create a look and saving as a 3D LUT is much more straight forward than using math to edit the code of a LUT. I understand that there are certain things in terms of moving colour values etc that using code is most likely more accurate for, but is it not possible to at least get a good LUT from the Da Vinci Resolve approach? Matthew Tomlinson the image science guru spoke about option 1 but not 2 and i'm wondering why this option did not come up since it is so much more straight forward for the average user?
Hi K. Wasley,
This is Matthew Tomlinson.
It is certainly reasonable to create your LUT via Da Vinci Resolve ( There are other software packages you could use as well to do the same thing i.e. Nuke, Baselight, Lustre). The main item to remember when doing so to create a LUT using option 2 is to remember that the LUT you are creating is tied to the display you are using. Meaning that if you create a LUT using Resolve using a monitor set to Rec709 Gamma 2.4 Legal Range and your LUT makes your test imagery look amazing, that LUT is tied to viewing an image on a Rec709 Gamma 2.4 Legal Range. This means that if your goal is to make a DCP of your movie, and you are using your LUT, that you will need to map your LUT from Rec709 Gamma 2.4 Legal range to XYZ. Which is actually not a big deal but the point is that Digital Cinema currently uses Dci-P3 colorspace. Which is a larger colorspace than Rec709.
I personally am a fan of creating a LUT that maps to P3 and then make derivatives to Rec709.
Starting from a Rec709 starting point is not a bad thing, just something to be aware of.
There really is nothing wrong with using your Option 2 but (as an Imaging Scientist), whenever I go this route I always scrutinize the resulting LUT to make sure that there is not clipping or clamping. I will also run extreme colors through the LUT to make sure there is no color saturation quantizing. (You can test this if you are using colored LED lights on set).
Please feel free to ask any questions that you may specifically have for your situation if I can help and please let me know if anything in my explanation is confusing or unclear.
Since it's clear the only way to see what you're doing is to view a display that doesn't use ICC color management technology is to get one of the new 4K/5K iMacs whose gamut is in Dci-P3 color space but I wonder if it will have to be set to a 2.4 gamma viewing condition or if display profile will have to be removed to turn of color management.
This is Matt, I am an Imaging Scientist that has had the honor of working with Roger on creating the LUT that he has used for certain shows.
The process of creating a LUT can be a very personal journey that includes the DP, Colorist and Imaging Science. I have had the privilege of creating custom LUTs for DP's and each one is specific to the needs of the that show. Point being, there is no real formula. It is about understanding what the DP is trying to achieve for that particular show and using test footage that the DP has shot to visual work through the process.
There is no magic formula or software package that you can lift from Roger's experiences to create your LUT. If you would like to create your LUT you should, I would recommend that you do so in colaboration with your colorist, and (if you have access) an Imaging Scientist. The goal is create the Look that you are trying to achieve.
One important note would be understand your goals. I am a big fan of creating show LUTs that are based on Digital Cinema and then create the rec709 version as a subset of the Digital Cinema version.
Roger has traditionally shot his material through his LUT, meaning that Roger lights on set while previewing his imagery.
What does that mean? It means that the LUT is not magic, it is a starting point base look. A large part of the magic comes from Roger and his ability to light a scene. Focus on lighting skills. Use that as your base point. A show LUT can be very useful but will never replace artistic lighting.
How you choose to use a LUT may not be how someone else uses a LUT in a show. I encourage you find your own path.
I understand my answer is a bit metaphoric and perhaps frustrating but my point is that the journey of creating a movie is a personal journey.
I, of course, am glad to discuss in greater detail or talk specifics about your situation if you would like.
Please let me know if there are any questions,
Hi Matthew Tomlinson,
Thanks so much for replying to my post, it is much appreciated. This whole LUT creation and image science field has been occupying my mind for some time. I am not naturally technical, but i would like to create a LUT that brings out a pleasingly natural 'cinematic' image. I am not happy with Arri's standard REC709. I have previously tweaked the REC709 through Arri's software, but it was really just to the extent of desaturating and adjusting the contrast rather than moving the values of different hues in their saturation or colour balance. I agree that lighting is critically important, but i also strongly feel that being able to start with a pleasing 'look' from the camera in terms of its contrast and colour reproduction is very important. I find a good LUT inspiring much like the way that film stock or a great grading 'look' is inspiring in comparison to an unpleasant digital colour space that some cameras out there will initially give you. I also feel that understanding a good LUT, much like a film stock, is important in the sense that it might effect decisions down the line. How saturated to make the paint on a wall, how much to warm up that light, how strong to go with the colours of costume and so on. I say this on the basis of trying to shoot as much in camera as possible rather than adjusting costume, paint or whatever it is in post.
Thank you for your invitation for further questions, i have some i would love to follow up with...
1. Isn't any LUT, whether created through option one or two as outlined in my previous post, subject to the monitor it was created/viewed on? I assume that even a LUT created from code would have to be viewed on a monitor and that monitor would have to be calibrated to a particular colour space? You specified that the monitor was important with option 2 in particular so i'm curious why this is not also equivalently the case when using option 1?
2. What are the restrictions of creating a LUT through something like Da Vinci Resolve? Can Da Vinci be just as accurate as a LUT created in code or with the proprietary software/process i presume you used to create Roger's LUT? Is a LUT created through coding perhaps more accurately able to move around and control the different elements of a colour? I thought perhaps Da Vinci could create broad strokes but perhaps not get into clinical detail? Did you see Steve Yedlin's (DoP of next Star Wars) display prep demo (http://www.yedlin.net/DisplayPrepDemo/») where he showed a test of film next to Alexa RAW engineered to look nearly identical? He used math and proprietary software that has taken years of research to formulate, so i assumed doing something like this in Da Vinci is not straight forward. Can a LUT created through code and one created through Da Vinci ultimately produce the same thing however complex?
3. Does Roger's LUT move colour hue's and saturation far from Arri's REC709 LUT base look? The colours of Roger's LUT feel natural so i'm curious how much the colours, broadly speaking, are being mapped to different values or saturation levels. Was the intention in the LUT creation process to look like film? I wouldn't ask you to break down the DNA of Roger's LUT, i understand that is proprietary and could actually hinder someone from finding their own 'look' but i'm interested in the process, however much you feel able to illuminate. For example did you start by just adding contrast and saturation to Arri's RAW file and then manipulate from that point? Do you find that adding contrast and saturation to Arri Raw is a good starting point?
4. You spoke about testing a LUT created through Da Vinci to see if it would clamp or clip colours. If a person is creating a 'natural' looking LUT is this problem much less likely to happen?
5. In general, how does a LUT respond to Arri ProRes XQ in comparison to Arri RAW in your experience?
6. What is the difference between XYZ and P3? As far as i understood XYZ replicates the eye and has to be conformed to P3 because that is the colour space of a cinema projector. Do you either essentially work in P3 (Cinema) or REC709 (TV) or are there further colour spaces that need to be considered?
7. There was mention that working via ICC colour management is a problem. Why is this? I have a monitor running directly out of my mac (i.e. via ICC) rather than a hardware device. I calibrated the monitor using a light measuring puck in conjunction with software to REC709 2.4. Is this not sufficient if the end display is also a REC709 2.4 TV or equivalent? I assumed that as long as my monitor is displaying the correct colour values and contrast for REC709 all should be ok.
Thank you so much for your time.
This 3 month old thread might clear some of your questions up...
Nice call back!
Hi K Wasley,
Let's take each question one at a time.
1) You are correct that a LUT is tied to a display. So you if you create a LUT while looking at a monitor that is set to Rec709, then that LUT is tied to Rec709. The key to creating a LUT that can be mapped to multiple displays/colorspaces is to create your look based on a larger color space and then map that larger space to smaller spaces. Meaning if you start with a P3 to Rec709 LUT and build your specific look just before the P3 to Rec709 LUT, you can preview your LUT on a Rec709 monitor but you have a LUT that is tied to P3 once you remove the P3 to Rec709 component. If you leave the P3 to rec709 component, that LUT is valid on a Rec709 display. You are correct that the display colorspace is valid in both option 1 and option 2.
2)Fantastic that you referenced Steve Yedlin, I have worked with him and helped in some of his experiments. He is extremely passionate about color and very technical in his process. The Divinci process is a broad stroke methodology. I present it as an option in that it may allow to create looks quickly with instantaneous feedback. If you want to take it to the next level, I encourage study. That is how Steve learned. Understand the differences between additive and subtractive color, understand film and how crosstalk works. To create the indepth look you may want to achieve I recommend understanding the differences between classic film stocks not only in grey axis but individual color saturation when mapping from black to white. To create the looks such as what Steve has been able to do, data collection is key.
3) The LUT/s that I have worked with Roger in creating have had nothing to do with the Arri LUTs. They are are completely independent and built with his direct guidance. In the general question of how to build a LUT. In general I like to keep color and contrast independent. Meaning that I would recommend first understanding the color pallet you want and then contrast. Then double check that color still works once contrast has been established.
4) Clipping and Clamping check is a technical check to make sure your LUT is smooth. "Natural" can mean different things so I'm really referring to the extreme brights and the extreme darks.
5) In my experience you can use the same LUT on both. As long as they are both mapped to the same colorspace, in this case LogCV3 Alexa Wide Gamut.
6) Currently Dci-P3 is the standard colorspace for Digital Cinema. XYZ is the colorspace that DCP's (DCDM's) are mapped to. XYZ is an extremely large colorspace that was created with future proofing in mind. In essence, Digital Cinema Projectors are P3 projectors. So when Digital Cinema is mapped to XYZ it's in essence P3 inside XYZ. The reason for this is that as Digital Projection improves and their colorspace capabilities improve, XYZ as the colorspace container for Digital Cinema will still be valid but larger colorspaces will be able to be mapped into them. Those future projectors will allow for current P3 projection to occur based on a calibration setting so we should always be able to see "intent". I would take a look at Rec2020. That's the up and coming colorspace and when people talk about HDR, this will be the medium that will employ that colorspace first.
7)Can you reference for me the ICC Profile issue? You can work with ICC Profiles. It's just a bit of a different system and to be honest, once Photoshop started accepting LUT's I haven't done a ton with with them. Usually ICC Profiles are used as a means of calibrating monitors, as the example you gave. Sometimes the hard part in utilizing them is that they can be integrated into an operating system and it's difficult to be certain of the order of operations that are occurring. Also, I try to avoid computers making assumptions for me when ever possible because I'm a bit of a control freak in that respect.
I hope that helps,
Please let me know if I missed the mark in any of my responses.
"7)...Also, I try to avoid computers making assumptions for me when ever possible because I'm a bit of a control freak in that respect."...quote from Matt regarding working with ICC profiles in a display space.
Not to avert you, Matt, from addressing K Wasley's understanding with even more mind numbing complexity, but just some clarity with regard to ICC profiles.
Seeing shooting digital cinema is basically still photography at 24 frames continuous burst rate per second at 1/24's shutter speed, why can't this industry get still and moving pictures operating on the same standards in order to future proof both formats to insure "intent" of the look is preserved.
ICC profiles with regard to ONLY display calibration employ the same color space transforms on the fly (matrices or LUTs to manipulate the video card on the computer) that function the same as the DCI-P3 to XYZ standards for projectors. Except with D-Cinema the device viewed on (projector) must always have to flip a switch by reading the DCP (I'm assuming) to do the conversion. That's a bit of an assumption on its own as well and seems to be working quite well from the current movies I've seen as long as everyone upstream & downstream in the production line adheres to a color space standard.
Or are current devices, smartphones, iPads, Blu-Ray, etc (not just projectors) the consumer views content on present the creator's intent close enough by adhering to a DCI standard? Is this standard easier to adhere to than ICC which uses a tagging of content data system that basically carries the converter switch within the tiff/jpeg with still images in the form of a profile? Video doesn't seem to be able to employ this color intent preserver strategy instead employs a standard color space describer and trust proprietary devices that play content stick to this standard.
It's a mess to say the least IMO.