More LUT Questions (3 replies and 8 comments)
I am attempting to create my own LUT to be used for on-camera monitoring, as well as a starting point in a grade. Not dissimilar to Roger's method. I have read many of the posts on this forum about LUT creation, and as user "K. Wasley" says in a post in this thread (), there seems to be two ways to create a LUT.
Math. Creating the LUT either from scratch or changing the math of an existing LUT. For example, the standard Arri LogC to Rec.709 LUT from Arri's website.
DI suite. Using a program such as Davinci Resolve to create the LUT. This of course does not allow for the finite control that math allows for, but is much more user friendly.
Roger claims his LUT is very simple, that it consists mainly of contrast and saturation, and that the look comes from the lighting and production design. I completely agree, and am not trying to take away from Roger's skill. But, I am not interested in Roger's LUT for its' "look." I'm more interested in the math behind it that makes it so versatile. The base of it is simple, but the way it was made (I assume) was not.
I have tried to create my own custom gamma curve within Resolve, even using editable splines for a smoother knee, but it is a very clumsy. It works in big strokes, with quick results-- perfect for a colorist, but not someone attempting to create a LUT. The biggest problem is consistency. If I create a LUT in Resolve, it looks excellent on one image, and ridiculous on the next.
So my question is, how is such a versatile LUT made? And how is the math controlled? There seems to be a big gap in available information between advanced Resolve techniques and Steve Yedlin-like math.
There just doesn't seem to be a lot of practical "how-to" advice on this.
Not Roger, but here's a quote from Roger concerning him developing his LUT.
"I developed my initial LUT by shooting a variety of images and then seeing how they translated with the LUT applied. Once they were translated as I had expected/hoped they would be, I settled on that LUT."
You mention that- "If I create a LUT in Resolve, it looks excellent on one image, and ridiculous on the next".
Assuming you are starting off with very well exposed images, did you try adjusting your LUT until all your images were satisfying? Also, there is a difference between an initial LUT and a final grade. Are you possibly skipping that step?
I'm sure other people on this site with far more knowledge than me will be able to help you, but one essential thing in creating a LUT is to use a professional calibrated monitor, ideally in a proper viewing environment. A regular computer monitor, no matter which model, is not sufficient.
That's exactly what I've been doing-- shooting a variety of images and testing it under various lighting conditions/situations. Inexperience and lack of a high-end professionally calibrated monitor definitely makes this difficult.
To answer your question, I am not skipping that middle step. When I said "Excellent in one image, and ridiculous in the next", perhaps I used too strong of a term. It is mainly a contrast problem. Very small things.
The main thing is that Resolve just doesn't have the level of control I am wanting. I find that using the curves produces somewhat muddy, almost milky blacks. Again, working in broad strokes. Compared to Roger's final images, or even the standard Alexa LUT, applied to the same image, which has crisp blacks. (This is not a problem with the "soft-clip" settings either)
That is why I'm interested in the math approach...
I'm relatively happy with where I'm at with it, but the little details persist to bug me.
Anyway, thank you very much for your response! The journey shall continue!
Not Roger but some initial thoughts.
You will be able to get crisp blacks etc no problem with resolves in built tools eg the curves and shadows + lift control. I would suggest as a starting point try to emulate Arri’s Rec 709 lut yourself within resolve. This is pretty easy and will just require dialling in the contrast and saturation which will bring a pretty close match. Test this on a LogC image.
A lot of LUTS in post houses such as efilm would have likely begun from a film emulation that will then have been tweaked using proprietary tools/software.
A good LUT will give you a good result across a wide range of images as long
as they have been exposed correctly. Ideally you want to collect eg 50 images shot in Arri logC that have all been exposed correctly and test your lut across all the images. You want your LUT to be fairly neutral, without a strong look. Ideally you would create your ‘look’ in camera with lighting rather than with a LUT.
I’ve personally bought LUTS from Cine grain pipeline. I used the Kodak 2383 emulation LUT and then tweaked this in resolve to my liking, eg reducing saturation and contrast, making greens more natural, pulling back blues etc. I’ve been using this LUT in the Alexa across a lot of projects and been pleased with the results. I did this because I didn’t really like Arri’s standard rec709 lut.
In terms of creating your own LUT you could just start with a LogC image, add contrast and saturation and then change any particular hues with the hue vs sat and hue vs hue control in resolve. Then test it across many other LogC images to see how it’s responding. If you want to get into more depth, you can start using fusion where you will have a greater control over contrast for example.
Steve Yedlin specifically wanted to emulate the look of neg to printed film with his LUT which I believe was Kodak 5218/9 printed to 2383. He did this by taking a skypanel, putting it in xy mode and running 6000 or so patches of varying colour and brightness levels across film stock. The film was then printed and using a spectrophotometer he measured the value of each of these patches. Then he did the same with the Alexa and using software/math modified the Alexa values to match the film values. This created a contrast and colour match with the print film stock. Obviously this process is very in depth and was undertaken to match a particular ‘look’, in this case film neg printed on film.
It’s a good idea to try to develop your taste for what you believe is a nice ‘look’. Eg how saturated you want your images to be, what colours you like and how bright you want them, what contrast you like. Arri’s Rec 709 lut for example has a little unpleasant bright and garish reds. Film on the other hand has deeper and darker reds when it saturates. As you develop your taste you can then create a lut that reflects what you find aesthetically pleasing. You will probably find that many LUTS cinematographers use are fairly natural and neutral without a strong stylised look to them, the stylised look is created via lighting and also via the colour correction. As I’m sure you’re aware if you read this forum, Roger in particular favours creating the ‘look’ with lighting primarily rather than in the grade.
I have a calibrated Flanders scientific monitor but also a lcd computer monitor. I find that when the lcd is calibrated via my i1 display probe to rec709 2.4 100nits it is very close to the Flanders (of course there are subtle but important differences in the Flanders ability to display the contrast range etc), so I would recommend doing that to at least get yourself in a good ballpark to judge from. The probe is pretty cheap.
Excellent response! I have used print film emulation-- whether it be in LUT form, or other. I find them to work well in some situations, but not others. I'm looking for a relatively neutral LUT as a starting point. (As you mentioned) Although, I do like the colors present in in film emulations, as they blend some of the hues and push everything warmer. Especially greens. (The greens on the LogC to 709 LUT are almost super-saturated...) But, I sometimes find them to be heavy-handed. It is an interesting idea to create a more neutral version of an emulation. You could then condense that down into LUT form for use on monitors, etc. I will try that. On the flip side, I know Roger's LUT doesn't try to emulate a film stock. It's very neutral. As you mentioned, he doesn't like to create the look in the grade. I'm not even sure if his LUT manipulates any hues. I guess it all depends on the cinematographer.
It seems the best way to go about doing this is just to experiment... and experiment some more. There's no getting around that.
Thank you very much for your response, it has been most helpful!!
Well, I could be completely wrong but I think it’s likely that Rogers LUT has it’s origins in profiled film data, especially In it’s colours. Many high end facilities will be basing their propriety LUTS on films colour and contrast response as a starting point. I’d also add that while you say Rogers LUT doesn’t shift hues I think if you look at his imagery carefully and compare it to just adding contrast and saturation to a LogC image you’ll find it’s pretty different. The colour response is different, it’s very different for example in comparison to arri’s standard rec709 lut. Therefore how colours saturate and the hue that they are has been manipulated. If you think about it, every LUT basically is shifting hues/colours somewhat. The LUT is taking a colour input and shifting it to a chosen colour output. The colour is being shifted in hue, saturation and luminance. If you look at for example the lips of people in Rogers LUT you’ll notice that they don’t appear like they do in Arri’s rec709 lut and this is because the colour of eg the reds/pinks has been manipulated inside the LUT, likely being more desaturated in comparison with the standard Arri lut. Also, look at the greens in Rogers LUT, much more natural than arri’s rec709 lut which as you point out are saturated and garish. Personally I think the LUT is really important and it’s kind of like a film stock. The LUT defines the hue, saturation and luminance of colours, along with the overall contrast of the image. If you have for example bright and garish reds and greens in your LUT it can have a profound impact on the feeling of an image. Try looking at reds that skew to a bright magenta compared with reds that when saturated reduce in luminance and shift more towards a darker blood colour in comparison. Big difference. Personally I much prefer the deeper darker blood reds than bright garish magenta reds. I feel I’m only just starting my journey of really looking at the colours I am seeing and developing a sensitivity and understanding for them and a taste for what I do and do not like. We’re so used to colour because we have seen it every day since we were born that I feel it takes a strange re-awakening to begin seeing them afresh and being able to analyse and understand what we are seeing, what we like and why. I think It’s pretty fascinating
Absolutely. There does seem to be something different about Roger's color. That's exactly what I was getting at with my original post. Of course, he likes to emphasize lighting over post manipulation, but as you said, choosing a LUT is similar to selecting a film stock, and plays an undeniable part in the final image. Its importance should not be diminished. On the other hand, in the second post in this thread (https://www.rogerdeakins.com/lighting-2/lut-for-lightning), he mentions not emulating a stock-- even going as far as calling it "bogus"... It is odd. My assumption would be that, while it doesn't emulate a print stock, it does have a specific hue mapping built in. Again, like you said, more natural greens, and softer reds.
Another thing is how the highlights behave. They roll off softly, but still maintain enough contrast. It takes a lot of tweaking on a lot of images to get it to that level of versatility.
Color study does bring up some interesting quandaries, doesn't it...
I think you're in good hands with K.Wasley's suggestions. One more setup point is - what color space are you delivering to?
DCI-P3 for theatrical release, REC.709 for internet/TV. Your monitor needs to be calibrated to one of these before you begin.
I would strongly suggest calibrating to rec709 2.4 gamma 100 nits. Many projects are graded on monitors calibrated to this and then they are converted to P3 for cinema projectors or there is a trim pass on a p3 calibrated projector. Not many colours of a typical scene are outside of Rec709. Most of the pointer gamut (gamut of all surface colours) are within 709, so your image typically does not contain a great amount of data which cannot be represented in 709. Unless you are working on high end theatrical films where the whole pipeline is P3 I think rec709 is the way to go.
I'm not delivering anything in P3 currently. Mainly Rec709. My monitor is in Rec709. So I'm okay with marrying the LUT to it.