Custom ACES Part 1: Building An ACES Workflow In A Resolve Node Tree

Joey this is an awesome approach! Thanks!
At the end it gets pretty creative and I am hoping Resolve 17 addresses your work arounds with Fusion clips and Title so we do not have to do all of these great hacks you are ferreting out.

Amazing tutorial, Joey! Would be ok for you to sandwich transformations to ACES LINEAR in gamma so as to put some LMT out of cct log encoding chain?

Thanks! I don’t see why not. I’m usually putting the ODT on the timeline level – and theres nothing stopping you from applying more nodes before it. You could use a ACES transform from ACEScct to linear, then whatever LMT you want, then from linear to whatever your final output is.

Thanks! yea I’ve been really enjoying working this way on a ton of different projects.
I do wish more tools were color-space aware, but in the world of OFX I can’t really even see how it would be possible for everybody to be on the same page. So I like to take a more flexible approach if possible!

Thanks for this Joey! I’ve been successfully using a similar pseudo ACES approach for a while now, and had never considered that the render cache format was causing clipping issues – very helpful info. Looking forward to part two to check our your ACES fixed node tree.

Joey, this was incredibly informative! One part that puzzled me, though, was how you handled the graphics. Once you grouped the graphics clips together and applied the transform into ACEScct at the group level, couldn’t you then simply put those clips under the original adjustment layer and use that one adjustment layer to handle all the ODT needs? Unless I’m missing something, the group-level transform effectively applied a sort of IDT to the graphics clips, didn’t it? So to go back to Rec709 (or any other output space), couldn’t you just set the transform on the upper adjustment clip to target Rec709, instead of disabling it, re-enabling the lower adjustment clip, and disabling the graphics-group transform? If so, didn’t the group-level transform negate the need for an adjustment that sits between the graded footage and the graphics? What am I missing?

Hey Alejandro –

The reason I did a transform into ACEScct on a group level was because the graphics themselves are not ACES-aware in the timeline. So yea – essentially I was just giving them an IDT so they would show correctly either under an ODT, or when rendering a linear GAM.

You absolutely could do it all with one ODT – the only reason I showed two was just to demonstrate that you could do one ODT for rec709, and then do a different one for a AP0 linear GAM if needed. Switching the ODT on a single adjustment layer (or timeline node) – would work just as well!

This was amazingly helpful to see along with examples for the problems as well. Thank you! Can’t wait for the next installment.

Thanks Joey, I like this best of both worlds approach, I really like how raw formats such as Red/R3d behave in ACES. Nice malleable curves and seemingly easier to get a decent base grade with clean colour separation.

What I’d love to see your approach to matching similar shots from the same scene in this aces scenario, ie. ive nailed the wide, lets try than on the mid and see if I can tweak a tiny bit, but get a different results in comparison to my normal workflow of printer light offset then colour space transform seems to. do you find it easier to work in the aces space for this or pre-transform, Do you work from the shadows up?

Joey,

Can I ask you a quality question? I fully understand that these transforms are mathematical and in theory have no affect on image quality.

I am doing a Red project right now. When I set up a timeline like this with Aces nodes (RWG to Aces CCT and then ACES CCT to .Rec 709) and then I set up the “normal way” with a simple color space transform (RWG to .Rec 709)….

The ACES image shows significant noise and banding and the normal method is far cleaner. Have you noticed this? Am I doing something wrong?

Printer lights work pretty well in ACES for most things, so a lot of the balance/adjustment i do is just offset inside the ACES space.
But the cool thing is if you want to do a printer point or two *before* the IDT – in this workflow you absolutely can.

Usually I’ll CST from whatever the camera original is, into LogC, work in LogC as needed and use the LogC IDT to get into ACES, and do the bulk of the grading there.

Interesting. It shouldn’t have any real quality difference. A few ideas –

-What are your red debayer settings? When you don’t use project-wide color management, you need to be sure the red clips are being debayered into the colorspace/gamma that you are using for your ACES transforms.

-On any of your color space transforms are you using tone mapping? when going between multiple log formats – tone mapping can go a bit crazy.

Any chance you could share some screengrabs of the settings/results?

Fantastic tutorial, Joey. Definitely will approach the aces workflow this way. Did a test yesterday and I love the way it feels. Feels organically like Rec709. I’ve been wanting to switch to an all aces workflow for my projects but just couldn’t find myself liking the sensitive of the controls. But after watching this tutorial and trying it for myself, I will definitely be using aces on my next project that starts in a few weeks. Thanks!

Hey Joey,

I’m running into something peculiar with this workflow, and I wanted to see if you know what’s going on here.

I’m able to verify (by comparing a screenshot taken in ACEScct with one taken in DaVinci YRGB with your IDT / ODT ACES Transform node set-up, as you did) that the initial images do indeed appear to be identical. However, adjustments that are applied to nodes in-between these nodes do not seem to produce the same effect in YRGB as they do in ACEScct.

For example, this is ARRI sample footage from an Alexa 65, within a project set-up for ACEScct (IDT – Alexa, ODT – Rec709, AP1 Color Science):

https://uploads.disquscdn.com/images/76670a25338e0a5fb7a4679d3173bbee6df05867b9c64b70204c7be77f3ce88b.jpg

And this is the same image within the same project, this time set-up for DaVinci YRGB with 2 ACES Transform node — one an IDT – Alexa node and the other an ODT – Rec709 node:

https://uploads.disquscdn.com/images/ecbb68f1efae6d66387f8640a46b16c756a53dbad0d338cae445668454fac5a3.jpg

Identical, awesome! I then go back to ACEScct, turn off the IDT and ODT nodes, add a 3rd node between them, and apply some quick adjustments to the image:

https://uploads.disquscdn.com/images/c25c451871e6304bc31b7638391a16c6dfd7b2fc0a80de52dbd8d25dcd5094b1.jpg

Switching back to DaVinci YRGB, I turn on the IDT and ODT nodes, and this is what the image looks like:

https://uploads.disquscdn.com/images/0c2089f1fcfceb8455ee3161b01ba7ebff25b9194673c6ace65fd8a1b4a90a48.jpg

Quite a big difference between the two. Is this to be expected? I’ve attempted to switch the Color Space and Gamma of the adjustment node to AP1 / ACEScct, but that was unsuccessful. Looking forward to hearing your thoughts!

Brilliant stuff. I have to admit, I’ve been skeptical and annoyed with ACES in the past, but this actually shuts up a lot of my objections to the process.

That’s really cool. Question:

Could you put the ODT in a shared node? That way you could place it anywhere (clips and groups) and change the ODT at once.

Absolutely!
I’ve had some issues with saving shared nodes in powergrades, which is why I personally haven’t done it with shared nodes – but yea a shared node would work perfectly for this,.

Thanks for beeing precise, on point and focused in your presentation 😉 very good.

I just made a rather interesting discovery. Perhaps this is a known fact, but, I never realized it before that shows my ignorance. First, I just discovered that by clicking on the 3 dots in the upper right hand corner of the Curves window(I’m using resolve v16.3 beta2), one can select “VIDEO” or “DATA”. By selecting “VIDEO”, the Curves display is remapped to show 16/235 IRE as an identifiable limit. The next discovery, made by showing “VIDEO” in Curves, is that ACES, when the ODT is set to REC709, will delimit all output to 16-235 IRE. Selecting sRGB, for example, in the ODT, will allow data display at 0-255 IRE. That is to say, it will clip the output at REC709 ranges, when the ODT is set to REC709..If one is grading “properly”, I suppose this is of little consequence. However, I was unaware that the ODT in REC709 mode, will clip all data outside of the spec range. This, of course, makes perfect sense. And it’s consistent with the “DATA” setting which shows 0-100% IRE. The only downside is that the waveform and histogram will not show how much data is exceeding the REC709 range limits.

I have been doing and recommending this workflow for a while now but recently decided to go back to ACEScc because I use my panels OFFSET mode a lot. Color balance and tint controls won’t work with custom aces as in YRGB those controls effect the gain and not the offset.
About the filmconvert and OFX stuff, I find that setting the nodes to a different gamma/gamut will mostly fix this for me. On the other hand, the node gamma/gamut won’t work in YRGB at all.
I would totally go back to custom ACES if it weren’t for those things. Have you found issues with this or maybe even a fix?

Hey Simon! Offset as in printer lights works great, but yea I can see how temp/tint wouldn’t work well.

Personally I never use those controls – I greatly prefer the chromatic adaptation OFX. I feel like it yields a much cleaner result, and you can specify the colorspace so it works both before the IDT in camera space, or after in ACES

I hadn’t even known about it. These days I went back to custom ACES and IDT to linear at first and use the new HDR tool to change WB in linear AP0 before CSTing to ACEScc for the rest of my workflow. No Offset wheels either but good results. I have yet to compare it to the Chromatic Adaption OFX.