starcentral
Veteran
Last edited:
F55: What ACES is
- Thread starter starcentral
- Start date
abreu-canedo
Well-known member
Thanks for the video!
Just a quick ACES test I did for a shoot. More info on the description section below the video on vimeo page.
https://vimeo.com/150983655
I found it to be very intuitive and while I am not the foremost expert on the matter, I am starting to enjoy the way ACES applies transforms as opposed to LUTs. Specially in S-gamut and wider color spaces. It has provided me with a great starting place to grade from.
Just a quick ACES test I did for a shoot. More info on the description section below the video on vimeo page.
https://vimeo.com/150983655
I found it to be very intuitive and while I am not the foremost expert on the matter, I am starting to enjoy the way ACES applies transforms as opposed to LUTs. Specially in S-gamut and wider color spaces. It has provided me with a great starting place to grade from.
paulcurtis
Veteran
ACES is awesome, but there's a lack of IDTs for a lot of cameras. Getting footage from RAW/ProPhoto and other spaces is a bit fiddly (go via XYZ into ACES) but can be done. If your camera and workflow is supported then as you say, it's super simple to get realistic results. The ACES RRT is a really nice transform to shape colours into a pleasing space without using print film LUTs and so on.
Also on your example shouldn't you be using an sRGB ODT for vimeo? Your blacks would be crushed a bit in 709 on a sRGB...
cheers
Paul
abreu-canedo
Well-known member
For sure, ACES is differentially supported amongst camera systems and it makes it harder for some systems as opposed to others. I use it on a F65, but have started to experiment with FS700/O7Q RAW.
I cannot comment on how well it works yet, since I have not done much with it, but I am pleasantly surprised. FS700 is advertised by sony to be wide color gamut, but it is not exactly S-gamut. I am trying to figure out a workflow for the FS700/O7Q where I am only using LUTs for output to Rec709 or P3DCI or sRGB and such. I just started doing ACES/CinemaDNG/... Still not sure what is going to work best after that.
Good observation, too! I used Rec709 because I wanted test for broadcast standard. Putting on vimeo was an afterthought.
Harry Lime
Well-known member
I've started to experimenting with ACES in Nuke and Resolve with F3 slog/sgamut1 footage and the results are probably the best I've seen out of the camera.
There is no IDT specific for the F3, so the only option is the F35 IDT. The footage come in a little on the red side and I'm pretty certain the F35 IDT is not a perfect match for the F3, but regardless it's a big improvement over what I've been getting previously.
There is no IDT specific for the F3, so the only option is the F35 IDT. The footage come in a little on the red side and I'm pretty certain the F35 IDT is not a perfect match for the F3, but regardless it's a big improvement over what I've been getting previously.
paulcurtis
Veteran
For sure, ACES is differentially supported amongst camera systems and it makes it harder for some systems as opposed to others. I use it on a F65, but have started to experiment with FS700/O7Q RAW.
I cannot comment on how well it works yet, since I have not done much with it, but I am pleasantly surprised. FS700 is advertised by sony to be wide color gamut, but it is not exactly S-gamut. I am trying to figure out a workflow for the FS700/O7Q where I am only using LUTs for output to Rec709 or P3DCI or sRGB and such. I just started doing ACES/CinemaDNG/... Still not sure what is going to work best after that.
Good observation, too! I used Rec709 because I wanted test for broadcast standard. Putting on vimeo was an afterthought.
I've been looking at taking RAW DNGs from the O7Q into Nuke (doing the colour work myself, because i can get XYZ space out of those RAWs, then from XYZ into ACES is a well defined route. I'm just experimenting with that now in fact and as you say the source RGB from the RAW is wider than 709.
Beyond that you can take any colorspace defined with Primaries into ACES via XYZ but then to truly define a camera *should* mean more than just primaries as a true spectral definition of a camera goes beyond that. But then again, in a lot of cases those primary definitions may be enough.
There is a cDNG IDT in Resolve but i'm not convinced whether it's any good, in my past experiences it didn't produce good results. Hence trying a different approach.
You will find in the DNGs from the O7Q definitions for the colour camera matrixes and you can use those, along with the neutral matrix to get to XYZ.
As i mentioned above you could move from Sgamut into ACES pretty easily? Whether the definition for the F35 is better is something you'd need to side by side. If you don't have it already Alex Fry has a nice gizmo that saves some time futzing with matrixes and finding the values for XYZ conversion. It's on his website.
There's no RAW from the F3 is there? So Slog/SGamut is the best you can have?
The beauty of ACES is partly the way it models the light and the RRT used, in some ways it's a nicer baseline that various emulation LUTs
cheers
Paul
Harry Lime
Well-known member
I'll take a look at that Gizmo. I'm still in the very early stages of playing around with ACES and NUKE.
The best the F3 can do is debayered uncompressed HD 10bit RGB 444 in slog1 / sgamut1 to an Odyssey (DPX). The F3 is one of those oddball HD broadcast - digital cinema camera hybrids like the F35/F23
So, far I'm really impressed with ACES. There may never be a proper IDT for the F3, but the future looks pretty bright after the past 10 years of color pipeline voodoo.
paulcurtis
Veteran
If you don't have already :
https://pro.sony.com/bbsccms/assets/files/mkt/cinema/solutions/slog_manual.pdf
Holds primaries for conversion to XYZ, once there you can convert to ACES primaries easily.
Scratch beneath the surface and colour science is voodoo, it's really really complicated and i've spent months wrapping my head around it, from first principals upwards and i'm still not 100% there, maybe never will be 100%. Diagrams like the CIExy space make it look simple, but those diagrams are really simplistic and don't really show what's going on.
ACES is a way to abstract a lot of that in a practical way, but it's also well worth understanding the pipeline from scratch. In nuke it's pretty easy to take a raw file (pre-debayer) and go through the steps of 'developing' it and see the matrixes being used and so on.
But true definition of an IDT should involve super secret manufacturer source, because 3 primaries isn't enough to really describe everything. Mind you looking through some IDTs it does seem that a lot of manufacturers do take the easy route. Or you have Arri who do the colour voodoo in camera to produce widegamut and that is defined with primaries.
But ACES is great for just getting decent images out, most of the time. ish.
In Nuke you have OCIO as well, and that is a good alternative to colour management. You've absorbed the cinematic colour papers?
cheers
Paul
Harry Lime
Well-known member
Thanks for that Paul.
I've been using the OCIO ACES setup in Nuke. It looks like you need to download the latest version of OCIO instead of using what comes standard with Nuke. This also involves setting some environment variables etc. Overall I wish the whole process to get it running was a lot less technical. I'm more of an 'operator' than a Nuke technical director / color scientist, so it's something of a PIA for me to make sure it's all running properly.
I agree that using the F35 IDT is less than ideal for the F3, but regardless the results still are better than anything I've personally seen before. Maybe Sony will throw us a bone and will release a proper F3 IDT, but I'm not going to hold my breath.
But so far ACES seems to work pretty well in Nuke, which is a good thing since I may use Nuke Studio instead of Resolve to grade my current project.
I've been using the OCIO ACES setup in Nuke. It looks like you need to download the latest version of OCIO instead of using what comes standard with Nuke. This also involves setting some environment variables etc. Overall I wish the whole process to get it running was a lot less technical. I'm more of an 'operator' than a Nuke technical director / color scientist, so it's something of a PIA for me to make sure it's all running properly.
I agree that using the F35 IDT is less than ideal for the F3, but regardless the results still are better than anything I've personally seen before. Maybe Sony will throw us a bone and will release a proper F3 IDT, but I'm not going to hold my breath.
But so far ACES seems to work pretty well in Nuke, which is a good thing since I may use Nuke Studio instead of Resolve to grade my current project.
Last edited:
SamEdwards
Well-known member
I did some tests with the FS5 slog3_sgamut3 to aces color transform in Flame. It gives a floating point image with incredible dynamic range, but seems a bit desaturated. I was also testing overexposure to keep the noise down on the camera, so I had to follow it with an exposure node to pull it back into range. It's also very contrasty and required a big bump down in contrast to look good. Further experimentation is coming.
A 4k image recorded over HDMI to the O7Q+ looked pretty darn clean and free of banding artifacts.
A 4k image recorded over HDMI to the O7Q+ looked pretty darn clean and free of banding artifacts.
SJX
Veteran
What? F35 is not hybrid broadcast camera. Is fully cinema camera. It even looks like film camera. F3 yes, 9000PL yes, but F35? No.
paulcurtis
Veteran
Actually after poking around with 0.7.1
Is seems that the IDT for the f35/f65 is basically a two step log to lin conversion (1D Shaper) and then a colour matrix. So in theory it would be pretty simple to adjust that for the F3 given the right starting point.
One thing to be wary of is that there have been so many slightly different versions of slog/slog2 etc,. If you read the comments in the .py script for the shaper LUT generation script you can see a small node to the pain of getting it right. Nuke has a discontinuity on that curve too if you look closely.
In addition to ACES it's worth looking at the spi-vfx OCIO setup, that's also pretty nice
cheers
Paul
starcentral
Veteran
To resume any subjects regarding ACES workflow.
Harry Lime
Well-known member
I say that because the image coming out of the F35 is in camera debayered HD. The same pipeline as the F3 etc. I'm not talking about how 'filmic looking' the output is but the workflow.
The F35 absolutely is a cinema camera in terms of functionality and the very filmic look it produces, but in terms of technology it is from the transition period when we went from shooting with HD broadcast cameras to hybrids like the F35 and ended up with RAW output cameras today that are debayered in post.
Last edited:
SJX
Veteran
In camera debayered HD? But F35 is CCD. Not CMOS bayer. I also don't understand where you see hybrid. F35 is HD cinema camera. It doesn't have to be RAW to be pure cinema camera. I don't see your point here sorry.
Harry Lime
Well-known member
It doesn't matter of it's a CCD or CMOS sensor. Both are bayer pattern sensors. The F35 sensor just happens to be striped.
If you look at the evolution of digital cinema cameras you will see that the beginnings are in broadcast video cameras. This goes all the way back to digi-beta and maybe D1 being transferred to film. But things got serious with the arrival HD broadcast cameras. I'm talking back in the days of the first Star Wars prequel which was shot on a modified SONY HD broadcast camera.
The next step in the evolution were cameras like the Viper and F35 / F23 etc. There were purpose built for cinema production and not just modified HD broadcast cameras. In many ways these cameras still had one foot in the world of HD broadcast technology, but also many features that we associate with a cine camera.
True digital cinema cameras really emerged with RAW recording that is debayered outside the camera. This step in the evolution allowed modern cinema cameras to shrug off the last of their broadcast heritage.
Don't get hung up on a purely technical classification. It's not intended as a criticism of the F35 which I am very fond of, but more to place it in the evolutionary timeline of the past 10-20 years.
Grug
Veteran
I'm not sure that's right. The bayer pattern is defined by being 50% Green, 25% Red, and 25% Blue. The striped array on the F35's CCD has equal amounts of each, so I think would fall into a separate category.
I'm racking my mind for the word that describes 'debayering' for non-bayer sensors though, and I'm pulling up blank on that front.
SJX
Veteran
This! Harry lime you know much about post production but I think you confusing this.