Hi, one question, where I work we make Netflix standard films, and their standard is Aces 2065-1, when I convert to Aces CG I don’t have flames, I have more saturated colors, would this for chroma key cropping help? Sometimes I convert aces 2065-1 to cct and I get good cuts, but not always.
1 Like
ACES2065-1 is absolutely not a good place to do pretty much any work. It is intended and designed strictly for transport between facilities and archiving.
Use ACEScg or ACEScct/cc for work, depending on the task/node in which you are working.
If you want more details you can check this out as it goes into tons of detail on what to work on in what color space.
1 Like
Ooh, I missed this in your original post…
This means something is incorrect with your Viewing Rules. An image is an image. ACES is designed for every image (captured by a known camera, that is) to look visually the same so we can composite them. BUT, under the hood when you look at the raw pixel data, it is in fact different encoding of pixel values.
Using the correct viewing rules means your ACES2065-1 and ACEScg should look the same, and only look different when you bypass the viewport monitor.
I watched your video, it helped me a lot, sometimes I have small doubts, but the ACES2065-1 even leaving it in bypass when viewing in flame has horrible colors, of course my work monitor is rec 709, cct the image I see washed log, for chroma key It helps a lot, but I’m going to try to use it to convert it into cg aces and at the end of the comp go back to ACES2065-1, as I have to give it to the colorist in ACES2065-1
Yes, a color difference only appears when I use bypass the view monitor
1 Like
Good, that means its behavior correctly. You are correct…when bypassing the Viewing Rules by hitting the Bypass button, you are looking at raw pixel data. To be clear, you really don’t need to be looking at the raw pixel data for working, only for double checking you know what you have is correct.
Use the Viewing Rules as I’ve outlined at the beginning of the series and you’ll be fine.
Good luck!
1 Like
Correct. ACES2065-1 (aka AP0) has different primaries from AP1 (cg, cc, cct, proxy)
If you don’t convert between the two, your color mapping (which pixel values represent which colors in CIE-XY will be factually incorrect).
But what Randy said is ‘if you use the correct viewing rules’, which refers to the matching transform, they will look the same, as they would apply the proper transform. So his statement is also correct with the ‘if’ he applied.
Meaning there would be a different transforms for AP0 and AP1.
As there’s not necessarily meta data, you may have to tag the files accordingly.
While it may sound pedestrian, I found that the ACES Wikipedia page is a very handy reference for ACES that is easier to decode than most way more technical documentation you can find: Academy Color Encoding System - Wikipedia
It has all the facts and no fluff. Scroll down to the ‘ACES Color Spaces’ section of the page.
1 Like
@mux Fair enough. I hadn’t read the whole thread from the top (should have).
On the original point - I think this would require more investigation. While technically a pipeline may work either way, and there’s a good argument for not throwing away precision if you don’t have to. There are three caveats, two of which are Flame specific:
- Are all the nodes working equally well in AP0 than in AP1 - things like keyers, etc? I believe I read that the origin of AP1 was to make it easier to work with some tools. That would require testing and comparison.
- Are the color spaces transforms/viewing rules that ship with Flame for AP0 as comprehensive as they are with AP1, or do you have to augment with your own? That would but a big burden, but is a one-time solve at the pipeline.
- Is the difference perceptual to warrant the effort. Kind of like the 13th bit debate on the Alexa35. And is the difference valuable to purists or the everyday practioner?
I’ve never seen AP0 as a working format.
Do you have to keep the bit depth at 32 bit? Do the clients notice if you don’t? Do they notice if you go into CCT and back? Do you have to give them a heads up if there are some ACEScg elements?
I’d probably just lie and work in aces CG.
Because at the end of the day, movies and TV aren’t good because they have colors that are nigh-invisible to the human eye. Nobody’s lamenting whatever colorspace Starship Troopers used.
But I mean, nobody listens to me. Instead the powers that be greenlight Rebel Moon and mandate an AP0 workflow. So I feel your pain, but I sincerely hope AP0 does not become industry standard.
2 Likes
Clients probably don’t notice but their QC dept that’s diff’ing everything with the originals at 3200% gain probably does.
2 Likes
definitely not the standard across most of my work but can confirm that I’ve had episodic projects come in that wanted everything AP0 top to bottom.
My question is do they ever do gamut kickbacks? Like “we’re seeing loss in extreme greens,” since that’s where the discrepancy between AP0 and AP1 resides.
Clipped whites are a universal issue in any linear colorspace and wouldn’t be affected by the ACEScg primaries vs the 2065-1 primaries.
And I know we’re all just working whatever way people pay us to work, but I’ve always been fascinated by extreme standards adherence (client side). From Netflix mandating sensor density, to what constitutes an “illegal color” in broadcast, to how goddamn big a “scanline” is, there is so much obfuscated jargon pushed by people who just heard it from someone else.
2 Likes
ah. not what I am saying.
All I was saying was an image encoded in different color spaces look identical when tagged correctly and viewed correctly.
2 Likes
I have never seen anywhere that a streamer will want you to work in ACES AP0. Where have you seen that?!! I work with all the big studios and I have never seen that recommendation ever, in fact, quite opposite. ACES even stipulates that ACES 2065-1 (AP0) is intended only as a transport colourspace and that you should only ever work in AP1 (ACEScg/ACEScc/ACEScct). I will actually even go as far as saying that if a major studio/streamer knew you were working in AP0 that they would have an issue with it.
There are good reasons why you don’t want to work in AP0. First of all, the gamut is way beyond anything any monitor can display. So you could be creating values you can’t see which will create clipping when you reach the extremities. Bad idea. A key could look good on your monitor now but down the track when Rec2020 capable displays actually become available you could have a whole lot of noise in them that you can’t see. Same goes for CG renders.
As for losing information, sure in Flame it is possible. Most other applications are actually working in 32bit so there is no loss in a colour space transform between AP0 to other colour spaces and back. The loss is so minimal in Flame that there would be no reason to avoid it.
Sure, if you want to go against all the technical advice against working in 2065-1 then go for it but it is not best practice.
3 Likes
Sorry, I call bullshit!
1 Like
Correct me if I’m wrong but 2065-1 is essentially like log color space, where acescg is more like traditional linear which all comp tools r written to work properly aka motion blur /glows/ add and screen mode etc
Produce a specs document then where they ask you to work in ACES 2065-1
Yep, when ACES, who developed the colour space distinction tell you NOT to use 2065-1 as a working colour space to then suggest they’re wrong is intriguing at best.
Because they aren’t.