ACES - A7r2 / s2

[ashleyBs]

Anyone experimented with ACES using the A7r2 or s2?
What program are you working with?
If so what IDT are you using?
What picture profile are you shooting in?
Lastly what are your findings and how do you like it?
Thanks

 

[paulcurtis]

I'm not using ACES with the A7sII but if i was this is how i would do it.

Camera to Slog2 tone curve and ProPhoto (the wider gamut space but any can work so long as you have the primaries)

Then depending what you're using ACES with you'll need to do something like

1. Linearise the tone curve - turn Slog2 back into scene linear
2. Transform the primaries of your chosen recording colorspace into the ACES primaries (being mindful of white point)

Those two steps are basically your poor mans IDT.

A *real* IDT could include a more accurate spectral response for the camera but then you'd be needing RAW output to take advantage of that.

The only reason to use ACES is really if you're working in an ACES pipeline overall. If you're only working with A7 footage in apps like PPro that don't do ACES then it's probably not worth your time. Instead i'd look into LUTCalc to produce some nice lots from the log source of the camera.

cheers
Paul

 

[ashleyBs]

Thanks for the reply.
Are you suggesting an ACES workflow does not work correctly outside of a raw workflow (from camera) or more that to use an IDT properly you need to be shooting RAW?
The software in my case is DaVinci and Nuke both of which support ACES, its a pipeline experiment vs any kind of LUT slapping for now.
This becomes a bit more of a resolve question but has anyone used the aces workflow with slog2 and aces in resolve?
Thanks

 

[Dermot]

i'm booked to grade a web series next week, and was testing the footage in ACEScc this evening as i normaly work with that, but i also normaly work with footage from Alexa/raw or Dragon.

I'm seeing a great starting point (and got to it very quckly) by using the Slog IDT with ACEScc in Resolve. I'm also not seeing any quantasion errors were clearly visable with any of the LUTs i have.

If i was going to work in display refered i would run far far away from any of the LUT's avb, they do nothing you cannot do in seconds with hands, eyes and scopes... and they do create issues , some create significant issues - some create less trouble, but all of them, every single on that i have create some damage to the image, utterly needless damage.

I agree wiith Paul, RAW workflow is preferable in ACES, log encoded footage has less to gain as it's already had the dynamic range compresed to match a display refrered workflow, where RAW from a camera capable of recording 16 stops has the most to gain by a scene refered workflow.

i use a three node basic set up, first node uses SOP+s, second node uses L*A*B*, third node uses RGB with luma mix set to 0, there i mainly work in primary bars, and almost never touch the hue offset balls, have all three of their sensitivity controls set to "1"

the choice of ACEScc for me is more about using the workflow and tools i'm comfortable with, and perhaps cleaner color seperation than it is about leaning on the imense highlight control you have with RAW in ACES

 

[paulcurtis]

Thanks for the reply.
Are you suggesting an ACES workflow does not work correctly outside of a raw workflow (from camera) or more that to use an IDT properly you need to be shooting RAW?
The software in my case is DaVinci and Nuke both of which support ACES, its a pipeline experiment vs any kind of LUT slapping for now.
This becomes a bit more of a resolve question but has anyone used the aces workflow with slog2 and aces in resolve?
Thanks

I'm really saying that if the rest of your pipeline is set up for ACES then it makes a lot of sense. But with this kind of footage it isn't a natural fit into that pipeline, it's quite generic. The IDT used won't 100% understand that cameras spectral response and the footage you are giving it has already been sampled into a simpler, smaller colorspace.

If you could get RAW out of it then the colorspace for RAW is quite a bit bigger than 709 but any movie output from it doesn't really appear to hold everything that the RAW could offer. Even ProPhoto which in theory is a larger space but on a practical basis not really. Again it's the Reds that tend to be out of gamut for sRGB/709 etc,.

There's no reason why Sony couldn't get us sensor RGB space or XYZ space out of this camera, that would be awesome and a simple firmware update.

All i'm saying is that you'd never loose anything going ACES and if your workflow already is then there's a workflow gain to be had. But you are only dealing with footage from this camera then you can do just as good a job with other tools and workflow - for example if you're in PPro and using Lumetri exclusively then that's all you need, there's no ACES there.

Compare to the new Varicam LT, is has ACES IDT and is designed around that workflow. The varicams produce wonderful colour images, beyond display referred spaces like 709, so ACES is essential in maintaining all the camera can see through the pipeline.

i'm booked to grade a web series next week, and was testing the footage in ACEScc this evening as i normaly work with that, but i also normaly work with footage from Alexa/raw or Dragon.

I'm seeing a great starting point (and got to it very quckly) by using the Slog IDT with ACEScc in Resolve. I'm also not seeing any quantasion errors were clearly visable with any of the LUTs i have.

If i was going to work in display refered i would run far far away from any of the LUT's avb, they do nothing you cannot do in seconds with hands, eyes and scopes... and they do create issues , some create significant issues - some create less trouble, but all of them, every single on that i have create some damage to the image, utterly needless damage.

I agree wiith Paul, RAW workflow is preferable in ACES, log encoded footage has less to gain as it's already had the dynamic range compresed to match a display refrered workflow, where RAW from a camera capable of recording 16 stops has the most to gain by a scene refered workflow.

i use a three node basic set up, first node uses SOP+s, second node uses L*A*B*, third node uses RGB with luma mix set to 0, there i mainly work in primary bars, and almost never touch the hue offset balls, have all three of their sensitivity controls set to "1"

the choice of ACEScc for me is more about using the workflow and tools i'm comfortable with, and perhaps cleaner color seperation than it is about leaning on the imense highlight control you have with RAW in ACES

+1 to all of this.

I would question what colour space that Slog IDT is expecting and whether what it is being fed is correct.

IMHO the only place LUTs count are either shaper LUTs or print emulation LUTs, the input and the very output. I'd prefer curves for 1D shapers but some cameras don't have simple curves, cough *red*. And film print is such a complex transform that you'd be hard pushed to mimic it from scratch.

The draw of ACES, apart from workflow, is that the RRT stage has a certain look. If you feed it HDR source (so linearised A7sII is HDR, Slog2 is HDR) then the RRT does a great job of compressing all that range into a 0 to 1 bucket for display.

If in a linear workflow you just stick a film print LUT at the end of the chain then what happens with all the values > 1? They get clipped. So there also has to be a shaper of some sort before a print LUT. The ACES RRT is a kind of print LUT but it's not a LUT.

I don't know if you have any tools to visualise a cube of colours and to see what kind of affect these different processes have on the footage?

cheers
Paul

 

[Dermot]

Paul, Ashley;

i'm just a grader, i have no idea what the camera settings are, no access to the DP, and no reply from production to my questions about camera settings.. it's clearly a log encoded something...

there's only two choices of IDT in Resolve, Slog2 & Slog3... i tried both, not a major difrence, went with Slog3.

I could test in Nucoda and Baselight, but this show is being conformed by production and they are handing me a Davinci project, so i'm gradeing Resolve.. and i've not sen any diffrences worth writing home about between ACEScc implemantation between the systems i use - huge diffrences between "Davinci ACES", "RCM" and "Truelight colorspaces", but nothing really noteable between ACEScc in diffrent systems

Have not had a need for a film emulation LUT in years, the few shows that still go out to print are mainly for south asia, and the labs i've worked with there prefer to use P3 10bit DPX file seq.

But i once reverse engineered the koji fuji film emulation lut in under 15 min just to prove a point.... it can be done, and done qucikly if there's a reason to do that... i think the "impossiable to recreate" is more marketing hype and pixel peeping than the reality of sitting it the hot seat all day long... any of them are pretty easy to reverse engineer to a point that MARCIE is visualy identical, the reality i see is that none of them are hard to match, and nothing but good will come from throwing canned LUTs in the trash as fast as possiable!

Nick Shaw has a DRX preset on offer to move Slog in 2.4 gamma without clipping, crushing or quantasion errors.. look for "antler power grades", he should have these on his website sometime this weekend;
www.antlerpost.com

 

[Nick_Shaw]

Nick Shaw has a DRX preset on offer to move Slog in 2.4 gamma without clipping, crushing or quantasion errors.. look for "antler power grades", he should have these on his website sometime this weekend;
www.antlerpost.com

Thanks Dermot.

Just to clarify, I have none of my recent stuff up on my website yet, and most likely won't do until the middle of next week. The Sony PowerGrade shown here is designed to match the LC709A LUT, but is a work in progress, and I wasn't planning to release that until I had done further testing. But if anybody desperately wants to buy it now they can contact me directly. It's the version which matches the standard Alexa LUT which is ready to buy now, and will be on the website next week.

Both the Sony and Arri PowerGrades are designed to match the stock manufacturers' LUTs but with the benefit that they don't clamp the output, and because they comprise a series of nodes which make up the transform, rather than a "black box" LUT, you can tweak the individual elements, rather than piling additional nodes on top of it to "fight" the look of the LUT.

 

[paulcurtis]

But i once reverse engineered the koji fuji film emulation lut in under 15 min just to prove a point.... it can be done, and done qucikly if there's a reason to do that... i think the "impossiable to recreate" is more marketing hype and pixel peeping than the reality of sitting it the hot seat all day long... any of them are pretty easy to reverse engineer to a point that MARCIE is visualy identical, the reality i see is that none of them are hard to match, and nothing but good will come from throwing canned LUTs in the trash as fast as possiable!

Ah but *you* can do that, most can't hence the market.

Thanks Dermot.

Just to clarify, I have none of my recent stuff up on my website yet, and most likely won't do until the middle of next week. The Sony PowerGrade shown here is designed to match the LC709A LUT, but is a work in progress, and I wasn't planning to release that until I had done further testing. But if anybody desperately wants to buy it now they can contact me directly. It's the version which matches the standard Alexa LUT which is ready to buy now, and will be on the website next week.

Both the Sony and Arri PowerGrades are designed to match the stock manufacturers' LUTs but with the benefit that they don't clamp the output, and because they comprise a series of nodes which make up the transform, rather than a "black box" LUT, you can tweak the individual elements, rather than piling additional nodes on top of it to "fight" the look of the LUT.

That's really interesting Nick, i've been messing with matrixes and spaces, helping get my head around stuff. I am surprised that the Arri 709A is achievable via matrixes, but that's a wonderful result.

I presume these are resolve only (that's a DRX?)?

cheers
Paul