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

[hyalinejim]

On 10/20/2023 at 3:54 PM, JulioD said:

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

If was able to nail exposure and white balance in every shot I wouldn't need any tweaking other than contrast.

But I don't think anyone expects there to be a magic transform that doesn't require some small bit of work. We also want to leave ourselves a bit of flexibility in deferring some decisions until post.

[JulioD]

5 hours ago, hyalinejim said:

If was able to nail exposure and white balance in every shot I wouldn't need any tweaking other than contrast.

But I don't think anyone expects there to be a magic transform that doesn't require some small bit of work. We also want to leave ourselves a bit of flexibility in deferring some decisions until post.

I dunno.  In this thread that’s been the complaint no?  One rule I took on from when I started way back shooting on film was to try to shoot the SAME exposure for each shot in a scene.  So don’t go changing it shot by shot, just have the same exposure once you’ve lit it the way you like.  Just set the exposure and don’t effin touch it again.  I often have colourists tell me how pleasant my work is to grade because it’s consistent.  Changes in post tend to be the same across shots then.

[kye]

2 hours ago, JulioD said:

I dunno.  In this thread that’s been the complaint no?  One rule I took on from when I started way back shooting on film was to try to shoot the SAME exposure for each shot in a scene.  So don’t go changing it shot by shot, just have the same exposure once you’ve lit it the way you like.  Just set the exposure and don’t effin touch it again.  I often have colourists tell me how pleasant my work is to grade because it’s consistent.  Changes in post tend to be the same across shots then.

It depends on the situation you're shooting in.  Obviously if you're in a studio and have complete control over the lighting etc then consistency is possible.

If you are shooting anywhere near natural light of any kind then there is always the possibility of lighting changes.  Even high-end productions shooting on days when there are only blue skies might end up shooting pickups later in the day when the light has changed a bit.  Anything shot anywhere near sunset is going to change shot-to-shot, and potentially even faster than you could adjust exposure and WB to compensate.

What is invisible on set might well be visible in post, considering how controlled the conditions are in the colouring studio and that you'd be cutting directly from one shot to the other.  

The reality is that with modern cameras the amount of leeway in post is absolutely incredible - latitude tests routinely show that you can push a shot 2 stops up or 3 / 4 / 5 stops down and it's still completely usable.  The idea that you'd try and shoot on set to require zero simple changes in post, which take literally a few seconds, is a false economy.

Here's one test showing Alexa overexposure tests: https://cinematography.net/alexa-over/alexa-skin-over.html

[JulioD]

@kye I think you missed the point. 
 

Just because you can over or under doesn’t mean it will grade the same way.  It’s not linear.  Especially once you get into a few stops over / under. 

If you want shot to shot consistency in grading maintaining the same exposure level helps a great deal. 

Of course you can recover a stop over or a stop under. But you’re going to have to park and grade it to match.  

Even in variable outdoor lighting it’s not too difficult to float your exposure using ND so that it’s very similar.  Notice I say alter your exposure to MAINTAIN the established exposure level / ratio.

That’s all Im saying. We seem to get very bogged down in individual shots and grades but chasing exposure shot to shot makes it so much harder to grade later on.

ETTR is a classic dead end in this regard in my view for MOTION work. It causes your exposure to roller coaster and while each individual shot may well look ok in-itself, you quickly find it can be impossible to make them all look good TOGETHER on a timeline. 

ETTR works great for photographers.   Not so great for cinematographers. 

[kye]

26 minutes ago, JulioD said:

@kye I think you missed the point. 
 

Just because you can over or under doesn’t mean it will grade the same way.  It’s not linear.  Especially once you get into a few stops over / under. 

If you want shot to shot consistency in grading maintaining the same exposure level helps a great deal. 

Of course you can recover a stop over or a stop under. But you’re going to have to park and grade it to match.  

Even in variable outdoor lighting it’s not too difficult to float your exposure using ND so that it’s very similar.  Notice I say alter your exposure to MAINTAIN the established exposure level / ratio.

That’s all Im saying. We seem to get very bogged down in individual shots and grades but chasing exposure shot to shot makes it so much harder to grade later on.

ETTR is a classic dead end in this regard in my view for MOTION work. It causes your exposure to roller coaster and while each individual shot may well look ok in-itself, you quickly find it can be impossible to make them all look good TOGETHER on a timeline. 

ETTR works great for photographers.   Not so great for cinematographers. 

I'm not sure what you're saying.

I shoot outdoors in completely uncontrolled available lighting, cinema verite style with zero direction and zero setup - I participate in events and I shoot as best as I can and I get what I get.  Matching exposure perfectly with zero adjustment in post under this situation is impossible.

Lots of folks shoot in a mix of available lighting with some control over events and so have some degree of control over the shooting, but not all.  Matching exposure perfectly with zero adjustment in post would be incredibly difficult and would come at the expense of other things that matter more, for example shooting fast enough to get additional material.

If you are able to shoot fast enough to completely and perfectly compensate for the changes in the environment you are in, and are able to do so to the tolerance of your taste in post then that's great, but it's a pipe dream for many.

Also, I think you radically underestimate what is possible in post if you have proper colour management setup.  I also think you potentially underestimate how small a variation it takes to benefit from a small tweak.  I don't see why a tiny tweak here or there is a big deal to be honest - even large changes are not visible on most material when done properly.

As such, just get shoot as best as you can, adjust in post as best as you can, and focus on being creative - almost everything else matters more to the final end-product than such a minute aspect of the image.

[kye]

36 minutes ago, JulioD said:

It’s not linear.

Actually, with proper colour management, it is Linear.

The Offset control in Resolve literally applies Linear gain, just like setting the node to Linear Gamma and then using the Gain wheel.

[hyalinejim]

16 minutes ago, JulioD said:

ETTR is a classic dead end in this regard in my view for MOTION work. It causes your exposure to roller coaster and while each individual shot may well look ok in-itself, you quickly find it can be impossible to make them all look good TOGETHER on a timeline. 

While I agree that it's very pleasant to grade a project where all clips have the same exposure level, I've found that I really like the ACES CC curve - because it's perfectly straight! This makes matching exposure a doddle. I just need to move the whole waveform up or down to match one shot with another. Again, if I could slap on a LUT and walk away I absolutely would - I would even burn it in!

But for real world shooting it's incredibly difficult to nail exposure every time. And by "nail" I mean get the exposure that looks best to me. The topic of how to expose correctly is a rabbit hole when shooting outside of the studio: angle of the grey card to the light source, how to use a reflective meter correctly, worrying about K constants etc. Even when shooting negative film, there is always room for adjusting the brightness of the image in the printing or scanning process to compensate for "errors" in exposure.

For me, with digital video, the ideal solution is to shoot log, adjust the signal in ACES CC space to balance all clips, and send that to a nice conversion. This is a slightly different approach to what @kye suggests in the first post, I think, which is to not accept a given conversion as is, but to grade each clip to taste after the conversion. Instead, I have a conversion that I feel is nice enough without further tweaking, but I am using ACES to digitally re-work the signal in such a way that it's very similar to what you would get if the scene were brighter/darker, high-key/low-key, warmer/cooler. In this way it's actually very similar to shooting RAW photographs and using a RAW converter. And it allows for conscious decisions such as ETTR strategies, as well as compensating for errors.

So my post work involves a brightness slider, a contrast slider and a colour wheel. And that's it!

[hyalinejim]

3 minutes ago, kye said:

Actually, with proper colour management, it is Linear.

Yes, exactly! If you can get your NLE to make linear adjustments to the data then corrections are a breeze. I force Premiere to do it using LUTs.

Log to ACES - corrections - ACES back to Log - Log to look LUT.

It's a clunky workaround, but it works!

[MrSMW]

1 hour ago, hyalinejim said:

So my post work involves a brightness slider, a contrast slider and a colour wheel. And that's it!

That’s about as much as I wish to do also.

Shoot for as consistent output as I can, then drop on a LUT that takes me most of the way and then fiddle with the exposure, temp, contrast, lows, mids, highs, add my subtle grade LUT, final tweak, done.

Having said that, the photographer in me is intrigued by the use of raw for video because maybe with 20+ years of experience shooting raw for stills… 

Or maybe raw video doesn’t work quite the same? 

[kye]

2 hours ago, hyalinejim said:

Yes, exactly! If you can get your NLE to make linear adjustments to the data then corrections are a breeze. I force Premiere to do it using LUTs.

Log to ACES - corrections - ACES back to Log - Log to look LUT.

It's a clunky workaround, but it works!

The more I think about colour grading in anything other than Resolve, the more difficult I see that it would be, and the more out-of-reach I can see that it is for normal people.

Good thing that the editor and other pages in Resolve are getting more and more suitable for doing large projects end-to-end...   just sayin' 😎😎😎

2 hours ago, hyalinejim said:

This is a slightly different approach to what @kye suggests in the first post, I think, which is to not accept a given conversion as is, but to grade each clip to taste after the conversion.

One thing I see many people being confused by is the order of colour grading transformations in the image pipeline vs the process of colour grading.

The pipeline that each clip should go on is:

1. conversion to a working colour space
2. clip-level adjustments
3. scene-level adjustments
4. overall "look" adjustments that apply to the whole timeline
5. conversion to 709 that also applies to the whole timeline

The process of colour grading is to setup 1 and 5, then 4, then 3, then finally implement 2, then export.

I have had thoughts about trying to create some tools that might be helpful and simplify things for folks, but between the barriers with using FCPX and PP, and the understanding that is required of pipeline vs process, I think it's too large a gap for me to be able to assist with.

1 hour ago, MrSMW said:

Or maybe raw video doesn’t work quite the same? 

RAW is RAW, regardless of how many of them you happen to capture.

[MrSMW]

1 hour ago, kye said:

RAW is RAW, regardless of how many of them you happen to capture.

I meant more along the workflow lines as in raw workflow with photography is to coin a phrase, a piece of piss to me as I have been doing it so long and I was wondering if because of that, I might find a raw video workflow something I could easily adapt to.

One day, I'll have a play...

[kye]

30 minutes ago, MrSMW said:

I meant more along the workflow lines as in raw workflow with photography is to coin a phrase, a piece of piss to me as I have been doing it so long and I was wondering if because of that, I might find a raw video workflow something I could easily adapt to.

One day, I'll have a play...

It's literally all the same, the only reason the software looks different is that in video there are things you want to do across multiple images, e.g. stabilisation.

You can colour grade video in Photoshop (I've done it, it literally exports a video file) and you can edit still images in Resolve.  

Every operation to an image that is done between the RAW file and the final deliverable is just math.

[hyalinejim]

3 hours ago, MrSMW said:

I meant more along the workflow lines as in raw workflow with photography

This is probably what Adobe were trying to do with the Lumetri colour panel, to provide Lightroom-like controls for video users. The only problem is that it can't tell what the input is, so its behaviour is nothing at all like Lightroom, unfortunately.

5 hours ago, kye said:

I have had thoughts about trying to create some tools that might be helpful and simplify things for folks

I'm not sure if this is related or not, but I remember there was a plugin called Logarist (which I never used). The idea was that you tell it what kind of log you're using and then its behaviour is consistent across cameras.

[kye]

15 hours ago, hyalinejim said:

This is probably what Adobe were trying to do with the Lumetri colour panel, to provide Lightroom-like controls for video users. The only problem is that it can't tell what the input is, so its behaviour is nothing at all like Lightroom, unfortunately.

I found LightRoom to be a very well-designed tool back when I was doing photos.  I read about wedding photographers ripping through several hundred images from a wedding and only needing seconds per one, and I could completely believe it.

Such an interface, where you just start at the top and go down through each section as required / desired, is a very easy experience.
I'm at a bit of a crossroads with my colour grading approach, where on the one hand I could implement a default node tree with heaps of nodes, and I could adjust different things in different nodes that are each configured in the right colour spaces etc, but the other pathway is for me to just design my own plugin that is like LightRoom and just has the sliders I want, each in the right colour space, that are in the order I want to apply them in.

15 hours ago, hyalinejim said:

I'm not sure if this is related or not, but I remember there was a plugin called Logarist (which I never used). The idea was that you tell it what kind of log you're using and then its behaviour is consistent across cameras.

I think this is why all those all-in-one solutions like Dehancer / Filmconvert / etc all have options to select the input colour space.  

The saving grace of all this is that most log profiles are so similar to each other that they mostly work with each other if you're willing to adjust a few sliders to compensate and aren't trying to really fine-tune a particular look.

If you don't have colour management (either the tool doesn't support it or you haven't learned it) then you're really fumbling in the dark.  To a certain extent I can understand people not wanting to learn it, because there are a few core concepts you need to wrap your head around, but on the other hand it is so incredibly powerful that it's kind of like being given a winning lottery ticket but not bothering to cash it in because you'd have to work out how to declare it on your next tax return.

[MrSMW]

1 hour ago, kye said:

I found LightRoom to be a very well-designed tool back when I was doing photos.  I read about wedding photographers ripping through several hundred images from a wedding and only needing seconds per one, and I could completely believe it.

Hello 👋 

[TomTheDP]

Somewhat unrelated to the topic. 

My attempt at matching an ARRI Alexa Classic and an FX30. 

Used an Xrite color checker passport chart. I guess that is cheating but I wanted to see if I could match them this way 100%. 

I think these are very close but maybe not 100%. Thoughts?

The more stylized one I switch the color space and gamma to RED and then put a Kodak lut on. The other two are just ARRI color space with the ARRI 709 lut. 

 

[kye]

1 hour ago, TomTheDP said:

Somewhat unrelated to the topic. 

My attempt at matching an ARRI Alexa Classic and an FX30. 

Used an Xrite color checker passport chart. I guess that is cheating but I wanted to see if I could match them this way 100%. 

I think these are very close but maybe not 100%. Thoughts?

The more stylized one I switch the color space and gamma to RED and then put a Kodak lut on. The other two are just ARRI color space with the ARRI 709 lut. 

 

Did you grade the FX30 to match the Alexa, or the other way around?

I see differences, but they're subtle and not significant.  IIRC it was Steve Yedlin that talked about the idea of the "cut back" test for camera matching.  Ie, if you were to have a shot from camera A, then another shot, then CUT BACK to camera B, would you notice that A and B were different.  
I think this is a good test because in real film-making you don't normally cut directly between two cameras showing the same scene from the same angle, so there is some distance there, and so in A/B testing you shouldn't cut directly between the same scene and angle because it's too stringent a test.
Yes, in theory you would cut between two angles of the same scene, but in lots of film-making you'd be changing the lighting setup between filming those angles, so getting a perfect match isn't relevant in these situations either.

[BrunoCH]

1 hour ago, TomTheDP said:

Somewhat unrelated to the topic. 

My attempt at matching an ARRI Alexa Classic and an FX30. 

Used an Xrite color checker passport chart. I guess that is cheating but I wanted to see if I could match them this way 100%. 

I think these are very close but maybe not 100%. Thoughts?

The more stylized one I switch the color space and gamma to RED and then put a Kodak lut on. The other two are just ARRI color space with the ARRI 709 lut. 

 

Very close. What jumps out at me is a difference in the blue of the sofa cushions in both and a difference in the saturation of the reds in the second still. 

[TomTheDP]

On 11/2/2023 at 7:27 PM, kye said:

Did you grade the FX30 to match the Alexa, or the other way around?

I see differences, but they're subtle and not significant.  IIRC it was Steve Yedlin that talked about the idea of the "cut back" test for camera matching.  Ie, if you were to have a shot from camera A, then another shot, then CUT BACK to camera B, would you notice that A and B were different.  
I think this is a good test because in real film-making you don't normally cut directly between two cameras showing the same scene from the same angle, so there is some distance there, and so in A/B testing you shouldn't cut directly between the same scene and angle because it's too stringent a test.
Yes, in theory you would cut between two angles of the same scene, but in lots of film-making you'd be changing the lighting setup between filming those angles, so getting a perfect match isn't relevant in these situations either.

That is very true. The only thing would be very interview situations where you have an A and B cam on the same subject. I haven't done one in almost a year now but I have a gig coming up where we are going to be doing that.

 

I used the color match tool in resolve, using an xrite chart. So you can set the color space and gamma, which I set to ARRI's. I did it on both cameras so it changed up the way the Alexa originally looked as well, with its preset white balance. 

This method would fall apart with creative lighting though. Let's say you wanted to do an orange tinted light as a key with blue for fill or something wacky. The chart is only useful for getting normal colors. 

 

On 11/2/2023 at 7:37 PM, BrunoCH said:

Very close. What jumps out at me is a difference in the blue of the sofa cushions in both and a difference in the saturation of the reds in the second still. 

Yes someone pointed both out to me. I didn't see them at first but its very apparent now. Definitely got it much closer than my manual attempts. I think with more fine tuning it could come closer. 

[kye]

2 hours ago, TomTheDP said:

That is very true. The only thing would be very interview situations where you have an A and B cam on the same subject. I haven't done one in almost a year now but I have a gig coming up where we are going to be doing that.

I used the color match tool in resolve, using an xrite chart. So you can set the color space and gamma, which I set to ARRI's. I did it on both cameras so it changed up the way the Alexa originally looked as well, with its preset white balance. 

This method would fall apart with creative lighting though. Let's say you wanted to do an orange tinted light as a key with blue for fill or something wacky. The chart is only useful for getting normal colors. 

Interesting.  It did a pretty good job then.

2 hours ago, TomTheDP said:

Yes someone pointed both out to me. I didn't see them at first but its very apparent now. Definitely got it much closer than my manual attempts. I think with more fine tuning it could come closer. 

The things that I noticed first was the skin tones.  This is what I always look at, and is most important.  The one on the left has those nutty tan and brown hues and the one on the right has 'porange' (pink&orange) and red tones.  The hues of the skin tones are the most important thing for me in images.

I also noticed the differences in the green/cyan area top-right, which has greens on the left but none on the right.  BUT, the skin tones are so important that they're the entire ballgame if you're cutting an interview between two angles.  

A really quick and dirty way to match skin tones when they're so far apart is to just rotate the hue of the entire image.  This sounds like a brutal thing to do to an image, but it will never break the image and we are so sensitive to skin tones that what is a huge change in skin tones is imperceptible on everything else in the image.  If there is anything that goes off in the background then you can easily do a Hue-vs curve adjustment on it, which is much more prone to inaccuracies and stressing the image so it's far preferable to rotate the hue of the whole image to suit the skin tones and then do the Hue curves on the background rather than the other way around.

This assumes that you've correctly white-balanced and exposed the cameras, and are doing colour management and transforms properly of course.