Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

ACES...

I find it very interesting to hear someone say that the ACES workflow makes their images look worse, because it seems to me that look really shouldn't enter into it. It should just be a workflow that can create any image that any other workflow could create. If there were technical limitations that could impact image quality, that would be one thing. But my understanding is that the bandwidth of 16-bit float should prevent that from being an issue. It'd be like saying TIFFs make your images look worse than PNG, or a Nikon NEF looks better than a DNG. They may have workflow advantages and disadvantages, but ultimately the same data is getting passed around.

Or maybe the problem is the RRT? If you don't like what it does to your images, then that could certainly be a problem. But you're not forced to use it are you?

I see IIF/ACES as just a modernization of the Cineon/DPX based workflows:

Cineon -> IIF (OpenEXR)
Log -> ACES (linear, with standard color primaries)
Film Look -> RRT


To me the main advantage is that instead of everyone passing around 10-bit log and a LUT, you bake it into a 16-bit float ACES file that everyone will read the same. Nothing should get lost in the translation, I should think. You could always convert the ACES back to log and work as you always have (and I'd expect many tools were doing just that).

But that's just me talking from a theoretical standpoint. Very interesting to hear real world use cases.


Brendan
 
They say... garbage in, garbage out. And therefore, quality in, quality out,
but a bad LUT or transform can ruin quality material.
 
The resulting look of an ACES rendered image is a combination of IDT (which we do, and is correct to spec) grading, RRT and ODT. If as in some of the R&D RRT+ODT combos the default "look" is too strong, it can look great with some shots and poor with others, as if it's heavily fine-tuned for one kind of image at the expense of others. It can also make grading to an "opposite" look rather hard. Can you imagine grading Girl With Dragon Tattoo look, if your view-LUT starting point gave you Pirates Of the Caribbean 4 as your starting point? Both excellent looks by excellent colourists, but both very very different....

Graeme
 
Horses for courses Graeme. Many of us depend on you for guidance here, the RAW paradigm is a complicated thing. All we want to do, is deliver to the client, in a way that makes the Epic shine.
 
The difference is that if you do an R3D based workflow you get to decide exactly what you want. With ACES... it decides the starting point. You work under that. And it is an aggressive starting point.

Jim
 
The ACES rendering does indeed define a starting point. That will influence the final image and the direction grading takes, and that's inevitable on how these things work. If ACES is really really neutral in it's rendering (which the currently released RRT/ODT combo is definitely not) then grading will be possible to practically any style or look you want, and the influence of the starting point will be minimal. I don't think ACES can meet that neutrality goal today with the software that testers have to use. We try really hard with our default renderings to keep neutrality in mind. A grey scale is grey without colour shift, for instance.


Graeme
 
I had the same experience of rendering through ACES workflow of same image from Arri Alexa - ProRes & ArriRaw. QT ProRes renders more saturated and biased skin tone where as ArriRaw renders clean image.

Definitely would say ACES is still in beta and its not advisable for production ready yet!
 
gbalaji said:
I had the same experience of rendering through ACES workflow of same image from Arri Alexa - ProRes & ArriRaw. QT ProRes renders more saturated and biased skin tone where as ArriRaw renders clean image.
Click to expand...

Not questioning your conclusion, but Arri Raw and the ProRes compressed images are identical in terms of color if identical Arri's LUTs are used during the RAW decode. I've done numerous tests on this to prove out that theory. When decoding Arri Raw, you specify the color space you're putting it into, and you also specify whether to embed the color matrix or not. The only way ACES would interpret them differently is if they were different to begin with, i.e., you embedded the color matrix in the ProRes version and not in the Raw version or vice versa. If that wasn't the case, there is no difference between the two provided you decoded the Arri RAW to Log C, and the ACES Alexa IDT wouldn't know they are from different original sources. My guess is that you either used different settings on the Raw decode than the ProRes files are created with, or you used something other than Arri software to do the RAW decode.
 
Jannard said:
The difference is that if you do an R3D based workflow you get to decide exactly what you want. With ACES... it decides the starting point. You work under that. And it is an aggressive starting point.

Jim
Click to expand...


Sometimes that's a very good thing. For some projects and some colorists, "cutting to the chase" can sometimes be welcome if - and only if - the chase you're cutting to is appropriate and desirable for that particular project. Although it might sound good to have things come out completely neutral, the truth is we don't always have time, especially in a client supervised session, to play around and explore different scenarios. You look at what's in front of you and go to where it seems to be leading you. The current RRT has some very specific characteristics that are quite nice for heavily saturated, mixed light environments, like some music videos and commercials, but inappropriate for others, like a "standard" drama that doesn't really scream out for a "look creation" scenario. It emphasizes orange/gold tones and cyan/blue tones in a considerably stronger way than most other color environments, and also increases contrast, particularly in the low end. It is, as you and Graeme say, not a "neutral" starting point. But for some projects, it can be useful, even if only to give the colorist some ideas.

Now, having said all that, I agree that it needs to have less of a "signature" and I and many others continue to make Alex, Ray, and others involved in the ACES project very aware of it. At this point it represents another choice. At some point in the future, it may or may not become more of a first choice. Time, and further development, will tell.
 
M Most said:
Sometimes that's a very good thing. For some projects and some colorists, "cutting to the chase" can sometimes be welcome if - and only if - the chase you're cutting to is appropriate and desirable for that particular project. Although it might sound good to have things come out completely neutral, the truth is we don't always have time, especially in a client supervised session, to play around and explore different scenarios. You look at what's in front of you and go to where it seems to be leading you. The current RRT has some very specific characteristics that are quite nice for heavily saturated, mixed light environments, like some music videos and commercials, but inappropriate for others, like a "standard" drama that doesn't really scream out for a "look creation" scenario. It emphasizes orange/gold tones and cyan/blue tones in a considerably stronger way than most other color environments, and also increases contrast, particularly in the low end. It is, as you and Graeme say, not a "neutral" starting point. But for some projects, it can be useful, even if only to give the colorist some ideas.

Now, having said all that, I agree that it needs to have less of a "signature" and I and many others continue to make Alex, Ray, and others involved in the ACES project very aware of it. At this point it represents another choice. At some point in the future, it may or may not become more of a first choice. Time, and further development, will tell.
Click to expand...


Not really. I tried the same source image from ProRes & ArriRaw and even send the sample to ACES developers to look into that. Since the feature film is yet to releasing in India, Cant send you the sample image for your reference.
 
Mike, I'm surprised to hear you say that in the context of what is being discussed here: the big picture, and not what might happen to fit the odd element of the big picture. I understand why, in practice, one might want to find a quick answer or direction to a specific need, but that's another story. Am I mistaken in thinking that a "neutral" starting point might also have quick avenues to certain looks? "Quick", and reliable, as far as the long term is concerned????
 
Bill,

I've never really fully bought into the desire for a neutral picture. As long the exposure is correct (not underexposed which creates shadow noise or overexposed which loses highlight information) I don't see how starting from neutral helps if you know how you want the picture to look from the outset. Neutral gives you possibilities in many directions. But if you know how you want the picture to look, I don't see how neutral is advantageous b/c you are exposing and recording for possibilities that you aren't going to use.

And actually, I could envison how neutral could be worse. For simplicity's sake, let's say that X amount of color correction causes Y amount of unwanted image distortion. If exposing and recording for the look you want means only 0.5X worth color correction, then you get only 0.5Y unwanted image distortion due to the correction. But if you expose and record neutral you may have to apply 3X worth of correction to move the image to the look originally intended for. That will add 3Y worth of unwanted image distortion.

Now the above is obviously simplistic... and more extreme grades acutally add more wonkiness to the image, but I think it illustrates the logic. Put another way, if I'm on Houston Street in Manhattan and I need to get to 14th Street, I take a cab to 14th Street which about a mile away. I don't ask the cabbie to take me to Central Park which is three miles away b/c it's in the middle of the City and the take me back down to 14th Street. If I need to drop-off stuff in Midtown or eat lunch on the Upper Westside, that would be helpful (ignoring the running meter, LOL!). But if I know where I want to go in the first place, it makes little sense to pass through netural.

I could be wrong about this, and it seems to go against a lot of the comments I hear colorist and other people say. So I'd love to hear Mike and others chime in on it! Thanks.
 
I hear you Peter. And Mike Most, for that matter. ( :

Is the issue with Aces a fundamental one: no matter how "neutral" or perfectly it interprets the RED data, it is still "interpreting", and that is seen as being inferior to the R3D workflow? And "Races" hints at a red-centic addition (damage control) to this less than perfect alternative to the R3D workflow?

At what point do we commit the data to the oven?
 
Peter, the types of image processing operations done by many of the RRT/ODT combos I've seen are not simple adjustments along the lines of what a colourist has access to in their traditional toolset. They may be theoretically invertible, but not practically so by colourist tools. That may mean that certain RRT/ODT combos may preclude certain looks because the standard toolset is not suitable for undoing their effect and to produce the desired look. Of course, if everything was a very simple linear transform, everything would be easily undoable with a simple toolset, which is my mind is almost another way of suggesting neutrality.

Graeme
 
Peter Moretti said:
Bill,

I've never really fully bought into the desire for a neutral picture. As long the exposure is correct (not underexposed which creates shadow noise or overexposed which loses highlight information) I don't see how starting from neutral helps if you know how you want the picture to look from the outset. Neutral gives you possibilities in many directions. But if you know how you want the picture to look, I don't see how neutral is advantageous b/c you are exposing and recording for possibilities that you aren't going to use.
Click to expand...

Neutral is important because you *know* what you're starting with. If I say "give me a perfectly linear image for compositing" and someone applied an unknown curve to the footage then you don't know how to get from the grade to Neutral. For example if someone takes an image and desaturates it... there is not returning to a color image. A Color -> B&W 3DLUT can't be reversed. A really contrasty high roll-off LUT from linear will compress lots of data into a very narrow data range. (Say for instance 3 stops of data in the data range 0.9 -> 1.0) Now whatever precision the data is stored in has inaccurate values for those 3 stops if you need to 'flatten' out the curve. So yes going from a neutral starting point to contrasty LUT does throw out a lot of data. But you're going to throw that out as meta-data or you're going to throw that out and it's going to be irreversible.
 
Peter Moretti said:
Bill,

I've never really fully bought into the desire for a neutral picture. As long the exposure is correct (not underexposed which creates shadow noise or overexposed which loses highlight information) I don't see how starting from neutral helps if you know how you want the picture to look from the outset. Neutral gives you possibilities in many directions. But if you know how you want the picture to look, I don't see how neutral is advantageous b/c you are exposing and recording for possibilities that you aren't going to use.

And actually, I could envison how neutral could be worse. For simplicity's sake, let's say that X amount of color correction causes Y amount of unwanted image distortion. If exposing and recording for the look you want means only 0.5X worth color correction, then you get only 0.5Y unwanted image distortion due to the correction. But if you expose and record neutral you may have to apply 3X worth of correction to move the image to the look originally intended for. That will add 3Y worth of unwanted image distortion.

Now the above is obviously simplistic... and more extreme grades acutally add more wonkiness to the image, but I think it illustrates the logic. Put another way, if I'm on Houston Street in Manhattan and I need to get to 14th Street, I take a cab to 14th Street which about a mile away. I don't ask the cabbie to take me to Central Park which is three miles away b/c it's in the middle of the City and the take me back down to 14th Street. If I need to drop-off stuff in Midtown or eat lunch on the Upper Westside, that would be helpful (ignoring the running meter, LOL!). But if I know where I want to go in the first place, it makes little sense to pass through netural.

I could be wrong about this, and it seems to go against a lot of the comments I hear colorist and other people say. So I'd love to hear Mike and others chime in on it! Thanks.
Click to expand...

i think you're confusing this with the whole 'shooting flat' dslr-craze and somewhat related workflows using log footage. that's a different discussion. there are legitimate reasons to keep things more 'neutral' (sometimes) when going to an rgb codec in order to preserve data - but that is for a different thread. if you're curious about that, post in ask mike most anything, or send me or somebody a pm. you could start your own thread - obviously, but i'm sure this has been discussed before, so search around a bit to see if that's necessary. we certainly don't need any excess/double threads clogging up poor reduser's arteries... :)

mike meant that working with aces should apply neutral transformations to your footage, meaning colors/contrast should not be affected. that's the whole problem with aces right now. you're not making the decisions, it's giving you a starting point. if that starting point is somewhere you'd probably go towards anyways... well, that's fine - otherwise it's a problem right now. the idea is not to provide another pre-packaged magic bullet looks type preset because you're working in aces - it's supposed to just be about making a widely supported standard that can be easily passed around. right now, it has it's own look - and it really shouldn't. i realize it's a difficult undertaking, but it should not be forced upon the entire industry until it is simple, effective, and 'neutral'. when it gets there, no force will be necessary.
 
An example of "not neutral" would be if you have an perfectly white balanced image of some neutral grey scale. If ACES rendered that with a warm tint, then it'd be "not neutral". In this case, a shift in white balance would be easily neutralized by any competent colourist.

Say the example is a bit more complex - a particular shade of blue is made purple. Just that shade of blue mind you, and no other. Could you construct a secondary to pull that shade of purple and make it blue? Sure, but you'd have to know such a thing was occurring to do it accurately. But what about if there's a shade of purple that is correctly portrayed, that might now get turned blue.... This kind of "not neutral" is much more concerning.

Graeme
 
You must log in or register to reply here.
Back
Top