Simple colour grading > Camera colour science (Why you should learn to colour grade)

[kye]

People seem to be obsessed with nit-picking the colour science of cameras, but indicate they don't colour grade for one reason or another.

To me, even a few simple adjustments can improve the image so much more than the differences in colour science between manufacturers.  In fact, the image out of camera is like a plain sponge cake straight out of the oven - it's nice and the quality matters but it's far from the final result.

Colour grading is also talked about as being super complicated, and it can be, but it doesn't have to be.  Simple grades can still be really powerful.

Here are some examples from online, to show how much of a nice image is camera colour science and how much is colour grading.

ARRI  LOG:

With ARRIs LUT:

Grade:

The above grade was done using only white balance, the lift / gamma / gain controls, a vignette, log wheels, in that order.

To look at skin tones, the holy grail of camera colour science - here's a before and after....   before:

Then on top of the previous look, here's additional treatments to give it more of a film look.

These additional adjustments were: Gain (to lower exposure), white balance, saturation (lowered), darken shadows, in that order.

Which was inspired by this frame from Sicario:

Here's the video showing the whole grade: https://youtu.be/8GkcqEA72QM

 

Next example - SOOC:

with 709 conversion:

Grade:

Video link: https://youtu.be/fRDjEB6ryyQ

 

Next one - with 709 conversion:

Grade:

Video: https://youtu.be/OmBBYHMi_ek

 

Next one - SOOC:

With 709 conversion: 

Grade:

Video: https://youtu.be/UNW_8jcGJqw

 

There are literally more examples online to count, but I just focused on the more neutral looking colour grades, as the people doing dirty film grades probably don't care about skin tone minutia when they're going to pummel the image with Dehancer etc.

So, what's the TLDR?

• Even half-a-dozen simple steps applied in addition to the manufacturers LUT can make a huge difference
• It's about making small changes to make the image look slightly nicer, and they add up
• The reason that fancy cameras look incredible is because the colourist takes the great work done on set and expands on it

How do I get started?

1. Look at the image and work out what tool might improve it (if you have no idea, just try the basic ones)
2. Wiggle whatever tool back and forth, deliberately going too far one way and too far the other way, then find the best spot
3. Compare the adjustment you just make to see if it makes the image better or worse, if it's better then keep it, otherwise undo
   (sometimes a really good adjustment will look completely natural and the 'before' will look like something is being applied to the image and is damaging it)
4. Go to 1.
5. Repeat until you can't find anything that makes the image look better.

If you're using another image to inspire your look, then for step 1, just look at both images and work out what looks different about yours, and try and fix it.  Is it brighter? Darker?  More contrasty?  A different colour?  More or less saturated?  Adding a vignette to lighten your subject or darken the other areas of the frame is a good trick.  Looking to find anything in the frame that's distracting and de-emphasising it is really useful too - even just lowering the brightness or saturation can really stop it from fighting for attention.

Even by the time you've adjusted these basic tools, you'll be well ahead.  

[PPNS]

just an interesting sidenote: basically every belgian dp agrees that the standard arri r709 conversion is way too tame. thats why they always shoot with a custom lut (usually per project)

[MrSMW]

1 hour ago, kye said:

People seem to be obsessed with nit-picking the colour science of cameras, but indicate they don't colour grade for one reason or another.

I think we are long since past the point where your average Sony/Canon/Nikon/Fuji/Whatever mirrorless is more than a few simple clicks or sliders away from being indistinguishable on-line.

I've seen so many side by side comparisons and can't see any difference at all such as in, "look how much better Y is compared with X!" and it looks near enough identical to me.

Same with the stills side. Waaaay too much emphasis on SOOC unedited files.

Comparing like for like across brands. Of course we can find an increasing proportional difference if we start comparing apples with kumquats...

Now things like difference in IBIS capability or wobble etc, some very clear differences, but modern sensors these days? Nah.

[markr041]

Leaving aside the condescending tone (really, we all need to learn color grading because we do not do it?), I agree with the OP that none of the default LUTs or transforms to REC709 are satisfactory and adjustments in post are almost always needed. And the examples are useful.

For the opportunistic video I shoot, for which it is impossible to set white balance manually and catch the shot, I use auto WB. And WB that is off is the number one problem with the ungraded but transformed clips. This is one reason why I shoot RAW, because WB adjustments are always needed. The examples given seem to be of studio shots with controlled lighting and time to set everything right in the camera, and so are good examples of the pure inadequacy of the transforms shot correctly (although one looks underexposed).

Generally, after transforms and WB adjustments I find saturation needs boosting as well as contrast, but skin tones are usually ok. For real RAW, sharpness needs tweaking too. The aim of course in my case is to mimic reality (the illusion of reality). None of this makes me an expert in color grading, and I am still learning some skills, but it has been enough to get positive "wows" from viewers seeing 8K 60 fps representations meant to convey they are there.

[JulioD]

I don’t find the tone condescending at all. 

And color correction can work for opportunistic shots just as well. Maybe even better.

I see so many LUTs being advertised packaged and sold and it’s actually not THAT hard to create your own unique individual look with only a little bit of effort.  
 

The thing no one seems to talk much about is MATCHING a look shot to shot. That’s where it gets difficult. And that’s where you are helped or hinder by the person shooting it. 

[MrSMW]

1 hour ago, markr041 said:

Leaving aside the condescending tone (really, we all need to learn color grading because we do not do it?), I agree with the OP that none of the default LUTs or transforms to REC709 are satisfactory and adjustments in post are almost always needed.

Not sure about the first part of your statement there Mark but agree with the second.

Lumix’s option for the S5ii and S1H is not great and I recently took a look at Nikon’s on the basis I might be going there for video (possibly) on top of the stills (definitely) and from the examples of theirs, it was horrendous.

I’m not sure why they don’t just pay someone with clearly far more skill than they do for this kind of thing as after all, what would it cost (very little) and we’re talking global releases here.

I’m sure if any brand popped out a stellar conversion LUT and a series of great conversion LUTS, they’d sell more cameras. Which is kind of the point is it not, to sell more stuff?!

I am interested in this topic, though not quite to the level of some, but will be exploring it more over my off season whether I stick with Lumix or go Nikon for my video needs. Especially if I go Nikon actually as I’d be starting from scratch and at a place I am mostly happy (or know what I need to do) with Lumix.

[JulioD]

It’s impossible to have a one size fits all LUT or else we would all be buying it. 
 

There’s so much variation in lighting and even sensor to sensor copy variation. Most people don’t realise that the color of a lens is usually DIFFERENT you shoot it wide open compared to three stops down. 
 

There’s no magic bullet for this. AI can maybe take over the color balancing at some point. Maybe. 

[kye]

13 hours ago, JulioD said:

I don’t find the tone condescending at all. 

And color correction can work for opportunistic shots just as well. Maybe even better.

I see so many LUTs being advertised packaged and sold and it’s actually not THAT hard to create your own unique individual look with only a little bit of effort.  

The thing no one seems to talk much about is MATCHING a look shot to shot. That’s where it gets difficult. And that’s where you are helped or hinder by the person shooting it. 

My experience with shooting auto-WB is that almost all shots definitely need some balancing done.  However, from all the (literally) hundreds of hours of colour grading stuff I've watched, it seems that every colourist always has a node for it, and talks about it like it's a thing they adjust on a regular basis.

The thing about matching shots is often not discussed.  Colour grading tutorials often just focus on grading one shot, so it makes sense that it's something that is sort-of implied too.

I had an epiphany when I was doing the Walter Volpatto Masterclass (you can still buy the recording - link) and he explained the whole process.

1. the entire production team design the overall look of the film and shoot and light each shot to have the right exposure and ratios (contrast)
2. the colourist then gets the footage and their job is to take that work and make it look good on the target displays by doing the following
3. applying a transform across the whole project to get the footage into the right display space (LUT, CST, etc)
4. working with the bosses to develop a creative look (modern, filmic, vintage, blockbuster, etc) and then applying that to the whole project
   (they might develop a few looks that are applied to parts of the film, eg, vintage look for flashbacks, etc, but it's mostly one look, and they should always work together)
5. performing minor adjustments to each clip to fine-tune elements like the WB or contrast (occasionally things change on set and you compensate for these, but mostly this stuff was determined by the camera and lighting on-set)
6. troubleshooting any issues with various shots..  for example, distracting objects in the background, dust on the lens, maybe some shots need noise reduction, etc.
7. then, only if there's time, would you look to finesse each shot and make it look glorious

As you see, the look goes over everything, and the shots are then adjusted as needed.  Often the colourist will apply the overall transform and look, then zoom out and just look at all the shots next to each other and looking for what stands out as being wrong or not fitting etc.  This way you can quickly bring the footage into alignment by just looking at the issues straight-up.  Then you start looking at each shot individually.  They also recommend working in passes, doing WB / contrast / sat on each for example, then doing power-windows etc.

This ensures that at any moment you have a film that is watchable.  If you finesse each shot in sequence, you might get half-way-through and run out of time and then you're completely screwed.

The way I do my projects (which are definitely NOT shot well) is to:

1. Apply my default nodes, which include the input transforms from each camera space, an overall timeline adjustment with 709 conversion and look nodes setup, and the default node tree on each shot
2. Adjust the overall look (choosing the transform, amount of subtractive saturation, etc etc)
3. Then I go through and exposure / WB / contrast / saturation for each shot
4. Then I go through and add power-windows to emphasise the subject(s) on each shot to direct the eye
5. Then I go through and remove any background elements that are distracting (too bright, too saturated, etc)
6. Then I fine-tune from there

It's a different workflow and different approach.

Despite shooting on full-auto, the cameras are getting better and they need less adjusting too.

12 hours ago, MrSMW said:

Not sure about the first part of your statement there Mark but agree with the second.

Lumix’s option for the S5ii and S1H is not great and I recently took a look at Nikon’s on the basis I might be going there for video (possibly) on top of the stills (definitely) and from the examples of theirs, it was horrendous.

I’m not sure why they don’t just pay someone with clearly far more skill than they do for this kind of thing as after all, what would it cost (very little) and we’re talking global releases here.

I’m sure if any brand popped out a stellar conversion LUT and a series of great conversion LUTS, they’d sell more cameras. Which is kind of the point is it not, to sell more stuff?!

I am interested in this topic, though not quite to the level of some, but will be exploring it more over my off season whether I stick with Lumix or go Nikon for my video needs. Especially if I go Nikon actually as I’d be starting from scratch and at a place I am mostly happy (or know what I need to do) with Lumix.

Not sure if you've come upon this before, but different countries / cultures often have wildly different colour preferences.  They also potentially have very different viewing equipment and viewing conditions.  At the same time that white people in the west were trying to look tanned, people across a significant proportion of the rest of the world were trying to look as light-skinned as possible, sometimes even risking their health with all sorts of skin bleaching and whitening products, and with darker skinned people often being discriminated against - colonialism really is the gift that won't stop giving.

In such situations, how do you create a LUT that positions skin tones for these completely opposing preferences?

Also, going back to the whole point of this thread - think about why the default transform isn't a final beautiful image..  it's not like manufacturers can't design a LUT with a strong look, YouTubers have no problems whatsoever doing it!  It's that they know that there is no universally desirable look, so they don't even bother to try.

Actually, that's partly incorrect.  They DO create finished looks, they're the 709 colour profiles in the camera.  It's assumed that if you're shooting LOG then you want to do it yourself.  

[MrSMW]

7 hours ago, JulioD said:

It’s impossible to have a one size fits all LUT or else we would all be buying it. 

Yes I know and agree. I simply mean a better starting point conversion LUT than most brands supply, ie, to get it in the ballpark in the first place.

For instance, in my experience, the conversion LUTs from Gamut and Phantom, are both waaaaay superior to the one Lumix provide.

I am just surprised that as this is an ongoing situation with every camera release, they don’t just do a deal with someone like Gamut or Phantom and have them do their base conversion LUT for them?

Ultimately I don’t care as I’ll go use the alternative and pay a few bucks for that but it surprises me the big boys don’t do better in house in the first place…

1 hour ago, kye said:

In such situations, how do you create a LUT that positions skin tones for these completely opposing preferences?

Yes I know and get that just as above, never can quite get my head round why most manufacturers basic conversion LUT’s are just a bit shit!

The reality is for every 1 of you Kye,,ie, someone who wants to DIY, there are 19 others who are not necessarily lazy and not expecting a single magic one click approach, but somewhere in between.

I have an outdoor (so much easier) almost one click approach nailed down but it’s the interior where mixed lighting mainly, is less consistent.

My approach this year has been:

Log outdoors, log indoors during daylight, flat profile in and out after dark/extreme low light.

I’m either going to go log outdoors in daylight + profile everywhere else, or back to profile all the time.

Consistency at the point of capture (as much as possible anyway and for most of my work, I have to work with the conditions I am given as either I don’t have the time to set up or control situations, or those situations are so short and varied, there isn’t any point anyway) and then in the edit as in of course I want my work to have a certain look, but within a certain timescale because you know, time is money!

[OleB]

9 hours ago, MrSMW said:

Yes I know and get that just as above, never can quite get my head round why most manufacturers basic conversion LUT’s are just a bit shit!

Before I started to develop the Rec709 conversion for the SIGMA fp I did quite a bit of research on how different manufacturers are creating theirs as a starting point.

One of the main ingredients seems to be that after the tone mapping of the cameras dynamic range into Rec709 nearly all manufacturers LUTs have a baked in default contrast around a pivot point of middle grey.

This looks great in a controlled studio scene, because it will be quite punchy out of the box. Additionally, it is hiding shadow noise because it will push these down as well.

However, in reality this means you are always pushing highlight information much higher by default than you would need to. Same goes for the shadow details. So usually the first post step would be to counter this effect.

My conclusion out of this was to create Rec709 LUTs, which are only tone mapping correctly and possibly create a smooth rolloff. Contrast has/can be added in post around a custom pivot point, or by adjusting the curves.

Let me demonstrate that effect on some sample stills, guess that is easier to understand. 

SIGMA fp Rec709 LUT (only colour and tone mapping into Rec709):

SIGMA fp Rec709 LUT + default contrast at a pivot of 45 IRE (intentionally a little bit too much to show the effect):

SIGMA fp Rec709 LUT + custom contrast at a pivot of 75 IRE

Guess this is one of the main reasons so many professional colourists are using CST/ACES workflows, to have a better starting point compared to manufacturer Rec709s. Maybe this information can provide useful for some of you. 🙂

 

[JulioD]

LUT’s and color are about TASTE.

Most conversion LUT’s are for a standard like REC 709 and guess what…they don’t look that great.

But they are accurate.

So they make a technically correct transform that accurately maps a larger than REC709 gamma and gamut from their sensor into a very limited TV ENGINEER STANDARD transform that is super accurate (to REC 709).  And even though we all say we want accurate color we really don’t, and not accurate color in a small restrained container like REC709.  

We complain.

Fuji have at least some other transforms they ship with their camera and that’s the right way to do it.

But if it’s a REC709 transform, it’s never going to look good.  It’s not meant to.

[kye]

12 hours ago, OleB said:

One of the main ingredients seems to be that after the tone mapping of the cameras dynamic range into Rec709 nearly all manufacturers LUTs have a baked in default contrast around a pivot point of middle grey.

My understanding that I have pieced together from snippets over many videos, articles, etc is that middle grey is important because it's where you would expose your skin tones, and then everything downstream of that makes sure to keep it in the middle and generally unmolested.

It's very common for elements of a colour grade to find middle grey, go down a bit from there, then make some adjustment that goes from that point down to zero (e.g. making the shadows cooler), or to go up from middle grey a bit then make an adjustment that goes up to the whites (e.g. highlight rolloffs, warm up the highlights, etc).

I don't think it's technically wrong to make an adjustment that messes with middle grey, but it's a convention that is useful. If you expose your material in this way, and everyone designs LUTs and curves and rolloffs and colour schemes etc that don't mess with it too much, then your skin tones will be preserved and come out the other end looking appropriate.  

It might seem a tedious and maybe OCD thing, but the opposite would be ridiculous - where no standard was the same and adjustments were all the Wild West and you would spend most of your time trying to find a series of adjustment that didn't push the skin tones down into the shadow region, then make them blue, then bring everything up so it's too high, then make everything slightly purple.  It would be a nightmare. 

2 hours ago, JulioD said:

LUT’s and color are about TASTE.

Most conversion LUT’s are for a standard like REC 709 and guess what…they don’t look that great.

But they are accurate.

So they make a technically correct transform that accurately maps a larger than REC709 gamma and gamut from their sensor into a very limited TV ENGINEER STANDARD transform that is super accurate (to REC 709).  And even though we all say we want accurate color we really don’t, and not accurate color in a small restrained container like REC709.  

We complain.

Fuji have at least some other transforms they ship with their camera and that’s the right way to do it.

But if it’s a REC709 transform, it’s never going to look good.  It’s not meant to.

Well said.

One thing I don't think younger people or non-professionals really understand is that up until very recently, video was completely dominated by a vast array of technical standards that were completely inflexible.  

I came to video from computers and the internet, so my perspective was that video is simply a series of images at a certain frame rate with a horizontal and vertical resolution with each pixel being a particular colour.  This is mostly true now, where you can have 1920x1080 at 24 fps but just as easily have 1922x1078 at 21fps.  But you don't have to go back in time very far before that wasn't the case.

Take for example the BM Ultrastudio Monitor 3G, a USB hardware device that works with Resolve to provide a signal for your reference monitor:

It has a TON of options for 2K:

Quote

2K Video Standards

2Kp23.98 DCI, 2Kp24 DCI, 2Kp25 DCI, 2Kp29.97 DCI, 2Kp30 DCI, 2Kp50 DCI, 2Kp59.94 DCI, 2Kp60 DCI
2KPsF23.98 DCI, 2KPsF24 DCI, 2KPsF25 DCI, 2KPsF29.97 DCI, 2KPsF30 DCI

First up, these are 2K video STANDARDS.  Also, note that there's 2Kp and 2KPsF - we all know that P means Progressive which means that each frame contains all the lines, which is different to Interlaced which is where each frame contains half the lines of the image and was a standard to sort of "fudge" extra resolution.  But PsF is "Progressive segmented Frame (PsF, sF, SF)" and "is a scheme designed to acquire, store, modify, and distribute progressive scan video using interlaced equipment."

If we look at the BM 4K Mini device that also does 4K outputs, we see that compatibility with tape machines is now no longer implemented for the UHD and 4K resolutions:

Quote

Ultra HD Video Standards

2160p23.98, 2160p24, 2160p25, 2160p29.97, 2160p30, 2160p50, 2160p59.94, 2160p60

4K Video Standards

4Kp23.98 DCI, 4Kp24 DCI, 4Kp25 DCI, 4Kp29.97 DCI, 4Kp30 DCI, 4Kp50 DCI, 4Kp59.94 DCI, 4Kp60 DCI

So, as recently as the creation of the 2K video standard, these hardware compatibility work-arounds were still in full-effect.  

If we got further back and look at SD standards, there are only two!

Quote

SD Video Standards

525i59.94 NTSC, 625i50 PAL

These are the only standards available, and the products don't even support outputting anything else.  

People who record digitally, edit digitally, and deliver digitally online, probably have no idea about the very recent past where video was about standards and technical compatibility.

The creativity that was created by the ability to colour video digitally - the Digital Intermediary (DI) process.  For reference, da Vinci 2K was the first system to support HD and 2K and was released in 1998 and cost $250,000 for the base price, with things like extra power windows costing extra (because they were additional hardware processing boards!).  

Before DI there basically wasn't any creative colour grading, beyond changing the way you treated the film by deliberately changing the exposure or colour casts of scenes or films.

This adherence to standards and conventions is not only still very present in the minds of manufacturers and production professionals, it's the entire framework for thinking about how TV and cinema is made.

The concept of "making a killer LUT" is as remote to the manufacturers as the idea of SMPTE or AESEBU standards compliance is to the LUT BROs of social media.  It's also why the industry players are so conventional in their thinking and have so much inertia, and why innovations from outside the industry is what is really shaking things up (like BM, Apple, etc).

[OleB]

2 hours ago, kye said:

My understanding that I have pieced together from snippets over many videos, articles, etc is that middle grey is important because it's where you would expose your skin tones, and then everything downstream of that makes sure to keep it in the middle and generally unmolested.

Yes, agree. That is why the tone mapping curve of my fp LUTs leaves this area untouched. However, this baked in contrast thing into most other Rec709 LUTs is in my eyes somewhat idiotic, as it is limiting the shown dynamic range of the camera already in the first step. Starting with the full dynamic in a more accessible way is my preferred starting point.

[MrSMW]

@OleB I've just started watching your series on the Sigma FP and it's really great what you have done and shared so big thumbs up from me.

I'm almost tempted with the FP (or L) but just one or two, 'too many' little quirks and issues that I can't quite get around for my needs.

Love the form factor with the Smallrig cage with wooden grip...which is as much as I'd want to rig it out and I know there is an aftermarket tilt option, but then I compare it to the S5ii and for my needs, sadly the Lumix 'wins'.

Sadly, because I'd switch to...or rather, would already be using the FP (or L).

Your series is still interesting however so thanks for that.

[hyalinejim]

On 10/18/2023 at 11:02 PM, JulioD said:

It’s impossible to have a one size fits all LUT or else we would all be buying it.

Lol! I have used just one LUT for every single client job I've done over the last 5 years.

It emulates the colour response of slightly overexposed Portra 400 film.

I pop it on and adjust brightness, contrast and white balance in ACES space before the LUT. It takes 15 to 30 seconds to do most clips, and makes matching shots a breeze.

I've made a new version for myself every time I get a new camera, so technically it hasn't been just one LUT, it's been three. But it does the same thing for each camera.

[MrSMW]

38 minutes ago, hyalinejim said:

Lol! I have used just one LUT for every single client job I've done over the last 5 years

Which reminds me... 😘

[OleB]

1 hour ago, MrSMW said:

I've just started watching your series on the Sigma FP and it's really great what you have done and shared so big thumbs up from me.

Thank you very much! 🙂 Glad to hear you find it valuable. 

[JulioD]

4 hours ago, hyalinejim said:

Lol! I have used just one LUT for every single client job I've done over the last 5 years.

It emulates the colour response of slightly overexposed Portra 400 film.

I pop it on and adjust brightness, contrast and white balance in ACES space before the LUT. It takes 15 to 30 seconds to do most clips, and makes matching shots a breeze.

I've made a new version for myself every time I get a new camera, so technically it hasn't been just one LUT, it's been three. But it does the same thing for each camera.

What you’re doing though is grading using a LUT as a shortcut to a look you’re trying to achieve.  
 

The point being there is no magic transform that just works without further tweaking. 
 

I bet too, if you had to grade footage I shot and lit it would react differently to the way you work because you’re grading your own material. 

[PannySVHS]

Is there a S35 mode for 12 CNDG Raw. @OleB One day I might like to try this beauty with a Pentax 28mm F2 in S35 cinema glory. Thanks for your and I think Liams findings and observational journey! 🙂

[OleB]

2 hours ago, PannySVHS said:

Is there a S35 mode for 12 CNDG Raw. @OleB One day I might like to try this beauty with a Pentax 28mm F2 in S35 cinema glory. Thanks for your and I think Liams findings and observational journey! 🙂

Pleasure! There is! 🙂 And it has got a faster readout speed compared to FF as well. I use it from time to time, if I am lazy to changes lenses.