LUT's, On set vs. DI (11 replies and 18 comments)
Dear Mr. Deakins,
You said that a Rec709 LUT is to color and contrast full in your taste. Now I know you use practically the same LUT on set, dailies and in DI; does this mean that the final look of your movies are more flat than movies shot by others? Beacause Rec709 is a delivery standard? Or am I totally wrong on this topic?
And for my understanding of you're workflow:
If you're LUT makes the final look to you're image, does that mean that besides 'color matching' and applying the LUT itself that is the most you do in the DI? For example, you don't add a subtle look to it?
Thanks a lot!
A little confusion here. Rec709 is a color space rather than a LUT and is, in my view, more suitable for television. I work in Log C color space with a proprietary LUT.
The LUT I use does transfer through to the DI suite but the LUT gives only a basic overall look. I still make changes to the images at the DI stage. I adjust contrast and saturation and I might window an area of a frame to bring it down or to reduce the saturation in that area. The DI is a powerful tool and is definitely an important part of the image making process these days but I don't believe, as some seem to, that the 'look' of a film is created in the DI suite. The 'look' of a film is created at the time of shooting and nothing that can be done in the DI can substitute for that.
I do and will totally agree that the 'look' is created on set!
The whole Rec709 space is confusing to me. But I never shot a real cinema production. I love the Log C space but I still have push some contrast and saturation to make it shine right? Log itself is simply to flat to deliver? If Rec709 is a 'television' space, is there a standard in between? Some sort of cinema space delivery protocol that I don't know about?
Thank you so mutch for sharing your DI workflow and helping me out on this one!!
I think it's about the balance of Log C and a good LUT. Log C by itself refers just to the color space.
But I am getting beyond my expertise and I will welcome someone who has more technical knowledge. I look at an image developed by a LUT and I ask for more contrast, less contrast, more saturation, less saturation. That's is the extent of my knowledge!
Thank you so much!
I Just found some clearness online. P3 is the color space that Efilm and many film post-houses finally deliver in. So my workflow is gonna be;
Shoot Log with my monitor Rec709 LUT > DI > Export to a Rec709 color space.
Shoot Log with my monitor P3 LUT > DI > Export to P3 color space.
Now the most of my projects are for online delivery. Does someone know which color space to use for online content? Are consumer displays like MacBooks and iPads capable for showing P3 space material or better stick to Rec709?
Online is all over the place... but I would harzard that Rec 709 is the most likely, as that is what many DSLRs are geared for, unless shooting in a raw format.
I tend to prefer something other than Rec 709, and so, the camera that I have shoots a raw format, so that's what I use for capture, then grade for a less than 'bright/saturated' color scheme.
The camera's i use are mostly: Sony PMW-F55, F5, FS7, Canon EOS C500 and the ARRI Amira. I always shoot LOG and mostly use a Rec709 space on set LUT.
I do not use DSLR's at all.
My question is more like what the final delivery color space is? For TV it is Rec709, and cinema mostly P3 which holds way more dynamic range than Rec709 does. But what is a Internet delivery standard? Is it useful to deliver P3, are consumer screens capable of showing this much information?
I think the defacto standard would be Rec 709. P3 would only be 'useful' if someone had some ability to display the media in P3 mode, which most internet users probably would not. They would use whatever display widget that came in their OS or VLC or the like, and if they wanted 'big screen' viewing use the Rec 709 compliant screen via an HDMI port.
Only users of the new iMac 4 and 5K screens are capable of viewing P3. The only thing probably is that the browser / player software also has to be P3 compatible.
P3 is a slightly bigger color space (which has nothing to do with dynamic range) -- what you'll mainly see between P3 and Rec.709 are things like how red is rendered (I believe it is a bit more orange in Rec.709). Technically Rec.709 and P3 are not labels for the gamma, and the standard gamma for Rec.709 monitors and P3 digital projectors are similar so the displayed dynamic range is similar. But the problem with many DSLR's is that they only record a dynamic range of 11 stops or so compared to the 14.5 stops of something like the Alexa in log or raw mode. This affects the flexibility of making corrections and the ability to compress overexposure information down within the display range of monitors and digital projectors. In terms of whether to build a display LUT (which is not recorded into the original, keep that in mind, it's just for viewing, often to correct a log signal for the gamma of a monitor) you'd make a Rec.709 LUT if your monitors on set are Rec.709, which most of them are. There are only a few monitors made that will display P3 color space. But again, we are talking about the color space, not the dynamic range.
Thanks you so much for your clean explanation! What still is a little bit confusing to me is the role of dynamic range and color space in LUTS, because a LOG file is about the DR, but does influence the color.. those are off course two separate things. So when I want more saturation and contrast in my LOG I use a Rec709 on set monitor LUT. Is contrast a fact of colorspace or DR?
And what would be the reason that Mr. Deakins is not recommending Rec709 LUTS. He sais the contrast and saturation are to high for his taste..
Thanks again, I appreciate your help a lot!
The standard Rec.709 viewing LUT for the Alexa that ARRI supplies with the camera is a bit too contrasty, saturated, and has a bit of a yellow cast in my opinion -- perhaps it's that particular LUT that Roger is referring to. But you can make a LUT for viewing on a Rec.709 monitor that looks like whatever you want -- desaturated, low-contrast, whatever. I think Roger is referring to what some cameras provide as a standard log-to-Rec.709 conversion LUT, which is why it can be useful to build your own. If you want to build log-to-P3 conversion LUT's, that's fine, if you plan on looking at the material using a P3 display device.
That must be it! So nothing wrong with Rec709 standard itself.
Wel, I probably only using Rec709 monitors on set. So a custom made LOG to 709 LUT would be fine! The only thing is when i'm using the same LUT in the DI I'm losing maybe more color space than I want to. But only for cinema projections.
Do you also are very LUT consistent like Roger is?
We do all our dailies and DI work at eFilm which is the only place to work in Log space, as I understand, which is important to us.
However, when we do the video master pass, we are working in Rec709 because the final product is going to be viewed on monitors. And we need another pass to bring the look back to what we set in LogC.
When we shoot on set, we have a specially calibrated monitor from eFilm to show us exactly what we are shooting, with our LUT applied on top.
LUTs are always on top of the recorded material and are not being baked in. If you are viewing a monitor, it makes sense to have a LUT designed for that color space.
A lot of DPs use different LUTs for different scenes. They set up the LUTs in the prep period. We do not work this way, we start with one LUT that is based on the LUT we created the first time we shot digital. On certain projects, we've gone in ahead of time to check the LUT with the specific movie in mind. For instance, a movie might want to have a bit more contrast or saturation. But it's basically the same LUT with tiny tweaks.
That probably comes from starting out in film where the emulsion was the emulsion and it was the lighting that determined how it looked.
I have reached out to an Image Science person with eFilm to do a few chats here on LUTs. He really knows the technical side but has an amazing ability to talk about it in terms that are easy to understand!
That would be amazing if the efilm guys could do a chat session it would be very helpful. I would love to know a step by step buy on creating a LUT. I am pretty lost on how to create a LUT from a LOG image.
This is Matt, the Imaging Science person James referred to. I have already responded another thread of yours that is very similar to this question. Perhaps could you start a thread that hits the specific questions you may have? I would like to help you and this topic is actually very large. So large that I am not even positive that a chat session would get us to where you would like to go. Perhaps if we could get ahead start before any chat session? Please let me know how I can help,
James!? Its such a lot of technical information from your side! Or was it Roger logged in at your account? 😉
But amazing, and thanks for de clear explanation. That you use one LUT for the entire movie instead of more for different scenes is exactly my workflow, so thats a nice confirmation to me! Do you know if Roger and eFilm output de LUT from the Alexa body to the SDI or is it applied 'in monitor'? And do you have a favorite monitor brand? I'm personally using TVLogic and Sony OLED.
I'm excited and happy to hear about your eFilm contact who might explain some more tech details.
Besides the standard Rec709 space LUT from ARRI (the one that Roger doesn't like) I use some film stock emulation LUT's that are great! But it would be very great to once create my own one. But I need a little bit more explanation about what a good starting point is. Thanks a lot for all the information and afford!
I am the technical geek of the two of us - so no, it wasn't Roger using my login. I oversee the digital workflow when we shoot.
I don't have time to answer right now but I do want you to know that Matt Tomilson the image science guru wants to become a member of the site and asked if he could be active and post in the forum. I said to him "Are you kidding me?????" So that will add to the fun.
To put it very simply: visually there's no difference between correctly displayed P3 and 709 footage. The only slight difference is that P3 is able to show slightly more saturated colours but since we don't often encounter such colours in daily life I don't think P3 is worth the effort and money for the small time cinematographers.
In theory there's a difference in white point and gamma value between P3 and 709 but the LUT of the monitor or projector corrects the image to what it is supposed to look like. Again: visually there's almost no difference, the untrained eye won't see much of a difference and you would have to shoot heavily saturated objects to be able to see it anyway. Of course if you have the money, P3 is the way to go because most facilities that do P3 will also have their monitor/projector calibrations (white point and gamma and luminance level) on point, which are the only 3 real important factors.
The reason why there's a difference between P3 and 709 in terms of gamma, luminance and white point is because of the fact that a dark theater and a (digital) projector is a very difference environment from a living room that has windows or practicals. Therefor gamma on P3 is around 2.4 in stead of 2.2 with 709 (2.2 is NOT a standard but a lot of people regard it as such. 2.4 has more contrast and this is evident because of the low light levels in a cinema: the eye adjusts and contast is lowered by your brain, so we have to increase the contrast on the projector to counter this effect. Second is the luminance level: when you display 100% white clip it should measure 14 fL (that's actually a standard) OR you could adjust the monitor/projector gain manually: the white screen should not be painful to your eyes, but it should be slightly "stingy".
The last thing is the whitepoint: yes there is a difference between 709 and P3 but again: the lut inside the monitor or projector corrects for the specific white point the footage is in so in the end -visually- there's not much difference if it is all done correctly aside from those slightly more saturated colours.
and white point is basically the same principle as color temperature balancing in a camera: you have to correct the source light of the monitor or projector so that the white point is actually white. Projectors need frequent calibration because the color temperature of the bulb changes dramatically over time due to aging.
David, when you say it appears to you that there's a difference between reds being a bit more orange I think what you refer to is actually metamerism failure between different panel or display technologies rather than the color space.. because in theory both P3 and 709 when calibrated properly should have exactly the same color rendering.
many people worry about monitoring but marketing makes it all very vague on purpose: to confuse as many people as possible so that you can make them believe they should pay copious amounts of money for all their equipment.
Any monitor that covers at least 95% of 709 and offers contrast, luminance and rgb adjustments inside will work very nicely: just make sure white is actually white by adjusting the RGB values and using any cheap calibration device. Make sure that the background behind your monitor is neutral, chroma-free grey and that you light it with a d65 light source with CRI of 90+
You'll adjust the gain and gamma value of the monitor in terms of the brightness of the room you're working in: depending on the ambient light, you have to set your gain in such a way that a white screen doesn't hurt your eyes but is also strong enough. You also need to adjust the gamma value. If you know that they advise most computer screens and tv's to be in 2.2 (considering people watch it in environments with a lot of ambient light) and on the other hand considering that in a completely dark environment they advise 2.4 I would suggest it is safe to say that using a value of 2.3 in a dimly lit room would be optimal! I've heard this from several other professional colorists that a lot of them use 2.3 because it is a nice balance between the two.
The reason your room has to be dimly lit is because if the room is completely dark you would have to turn the gain of the monitor very low.. The problem with this is that most consumer screens lose a lot of contrast ratio when you dim the screen that much: the blacks will become gray and the images appears a bit more washed out and it is more difficult to see detail in the picture. By increasing the ambient light with color correct d65 light you can also increase the luminance of the monitor and thereby you can get a much better contrast ratio!
Hello, This is Matt, the Imaging Science Guru that James mentioned.
I will do my best to respond. In reality I am seeing multiple questions being asked, please let me know if I am not clear or if I am misinterpreting the questions.
First I believe there is a misunderstanding between what a Lut is and what colorspace is. A Lut, expects to be fed one colorspace and it maps to another colorspace. For this discussion I will use the Alexa camera as the example camera. The Alexa is sending a LogC, Alexa Wide Gamut image to the Lut that Roger and James use. What happens in that Lut is the customization that the Imaging Science team at EFilm were able to provide, the final step in the Lut maps the image to the desired colorspace. This means that the Lut that Roger and James use is always specifically paired with the display colorspace that they are viewing images on. By doing so the imagery always looks as expected.
Colorspace refers to color gamut available.
Log is not a colorspace, Log refers to a tone map. LogC refers to a tone map. Slog2, Slog3 refer to a tone map. One could consider this to be contrast or gamma but in reality it is more complicated because it deals with the shoulder rolloff, toe rolloff and midtones. All of which can be customized. Common vernacular infers the colorspace that is associated with the tone map. Usually when people say "Log" they mean film log. When people say LogC, they mean LogC Alexa Wide Gamut, when people say Slog3.Cine, they mean Slog3, SGamut3.Cine. That is the tone map and the colormap.
So you can create a Lut that expects to be fed LogC Alexa Wide Gamut and maps to your desired color space, which includes a starting point look that can be expanded upon within the DI, as James mentioned.
There was the question of what to do with internet distrabution. For that I would recommend researching and testing the formatting that is required by the site you will be posting on. I agree that pretty much the safe bet is Rec709. The standard gamma for gamma for Rec709 (SMPTE) is 2.4. One concept to be very aware of is whether the imagery going online should be Extended or Legal (Broadcast) range. If you feed the provider what they need it should fit into their system nicely. Testing is always a good idea. The unfortunate aspect is that there is no guarantee that person who is watching your movie on their computer, phone, tablet, TV, is set to a standard. The strongest move is make sure that what you send is set the standard expected.
Please let me know if there are topics within this discussion that I have missed or you would like expanded upon.
I thought there wasn't really a gamma standard since the perceived contrast of a display is determined heavily by the ambient light in the environment and the reaction(s) of our eyes to it? I thought 2.2 was considered the "standard" on the web (and photography sRGB etc) and 2.4 in the cinema which is why I read and hear a lot of colorists monitoring in 2.3 and in dimly but lit environments because this is appears to be the best balance for each 'situation'.
For instance: gamma 2.4 appears crushed and contrasty when viewed in an environment with some ambient light hitting the screen and our eyes.
gamma 2.2 appears washed out and flat when viewed in a completely dark room.
Contrast is incredibly hard to get right because it depends so much on the environment which is why I don't think there is a specific standard.. I believe I've read or heard of some formula's that take the measured ambient light level in the room to calculate precisely the correct gamma value.
Question: in what sense is the LUT that you've developed with Roger and James different from the standard Arri mapping? I mean, the colors you get from the Arri LUT are pretty much spot on and look very natural, no? Or is there something wrong with it, you reckon?
Or is it only small adjustments in the luma curve that you do?
For a long time SMPTE had not set a standard for gamma for Rec709 and that did cause a lot of chaos. Recently (about 2 years ago?) SMPTE did release that the standard for Rec709 is gamma 2.4. sRGB is a different colorspace and to get overtly technical on it is basically a "perceived" gamma 2.2 because the flat end at the bottom.
The standard for Digital Cinema is gamma 2.6, based on the Dci-P3 spec.
Something to think about. Broadcast Levels (i.e. Legal or Headspace or 7.5 IRE) or Extended Range (i.e. Foolscap or 0 IRE). The way you described the difference between Rec709 and sRGB is often times the description I will receive when there is a Headspace/Fullspace mismatch. Just something to be aware of.
The Arri options of the K1,2,3 S1,2,3 are very valid LUT options and I personally have used them on shows. Roger and James have a LUT that has been customized for Roger. Neither one is wrong. They are simply different.
Please let me know if there are any questions,
Matt, thanks a lot for contributing in this forum. I have a question for your. It is more of an opinion.
First of all, I do not know if you have worked with Slog3 SGamut3.cine and its LC709A lut. I imagine that you must use your own luts. But in any case, when I paired Slog3 SGamut3.cine and the LC709A lut, I did not like too much the resulting colours. There was a strange vibrance in them, especially in yellows and greens, and reds seemed to shift slightly towards magenta. I was not able to reproduce with that lut what my eyes had seen colour wise during filming.
So I made a test of using the LC709 output gamma, but combined with the the Alexa Wide Gamut, to ake a new lut (this combination was made through the site Lut Calculator). The result was that the images had a very good contrast to start with, albeit very little saturation. But by increasing saturation, especially in the midtones and a little in the highlights, I found the colours to be much more faithful and beautiful.
What would you think of this approach? I know that for a more precise answer you would need to make tests, but in theory, do you consider it a valid approach? Or would it be a "technical crime" to output the Slog3 SGamut3.cine with the Alexa Wide Gamut and increased saturation? And if by any chance you have worked with Slog3 SGamut3.cine, could you tell me a little bit about your experience with it? Would you have a suggestion on how to work better its colours?
RAS, EFILM as a service provider does not necessarily pick the camera, the DP does.
And cause of this we worked with all of these cameras.
We do sign NDAs with all camera manufactures and they give us their white paper describing color primaries, curves and characteristics.
That way we are able to build our own input transforms and can match one camera to another. (as good as possible … we are all aware that there are differences)
To go from a Log camera space to a monitoring Rec709 or P3 viewing space we do need an S-Curve.
The Log signal coming in, combined with the viewing LUT will give you this S-Curve characteristic.
And yes mixing 2x or 3x LUTs is a good way to go. In our EFILM library we do have all camera manufactures LUTs available (including the Sony LC-709 and LC-709A mentioned in your blog). Additionally we have all characterizations of Kodak, Fuji and Agfa print and neg stocks. We do mix and match our LUTs as needed.
Our colorist starts with a verbal discussion with the DP and the imaging scientist builds a couple of LUTs based on the instructions of Colorist and DP.
In their next color grading session the DP and the colorist will pick a show LUT. This LUT will be used by the DIT on set, during dailies, during the trailers and VFX work and it will be used in the DI.
JZ, thanks a lot for such a detailed and helpful answer! It really clears things up for me. I will make some new tests based on you comments.
It is a privilege to be able to receive advice and insights from professionals like you. And I am also very grateful to Roger and James for the opportunity they give us.
Hi everyone! Really interesting thread.
I have a request that might be to much to ask but I'm asking anyway.
Something that would be really interesting to see is an ungraded frame grab from one of Rogers digitally shot films.
It would be super cool to see the difference and really understand what happens in a professional grade.
I totally understand if this is a little bit to personal to show or if the copyrights makes it impossible.
Just had to ask... Thank you for the amazing new forum!!