Color Space Transform vs RCM vs ACES

Which is best for color grading? Frequently, I am just using YRGB with color space transform since I have many cameras shooting in different color spaces.
Is there a right way to do it?

I have...

(school provided) Red Scarlet shooting RedCode3 or whatever it will shoot (it's raw so I don't think that matters)

Panasonic S5 shooting V-Log/V-Gamut internal or (any) when shooting BRaw
(or CineD2/R709 when shooting >51200 ISO)

Sony A7RIV shooting HLG3/R.2020 and/or S-Log2/ITU.709 Matrix (thanks 8-bit)

FujiFilm X-T4 shooting F-Log/R.2020

GoPro Hero9 shooting Flat/R.709?

iPhone 12 Pro Max shooting HLG/R.2020 using Filmic Pro

Color space transform I see as a good, quick way to do things, but I don't know how accurate it is.

RCM and ACES I have no idea what the difference is other than them being so confusing that they're not worth it unless they'll make my image quality better then CST.

I am a sub-1000 subscriber YouTuber for reference. Not a professional but very technically-minded.

going from vlog to rec709 with a color space transform node does exactly the same thing as doing it the color management way. However, doing it all with nodes, groups, etc, i found that it gives more flexibility and gives me more control over things like graphics, text, etc. Last time i used color management i encountered some fancy behaviour with those elements, although i've read in a recent update log that things were improved on that side. ACES in super short is just a different color space (like davinci wide gamut) and has a slightly different look after the rec709 conversion than if you do it with the resolve flavor. It's a matter of personal preference. ACES is also more of a standard worldwide, as it is adopted in all kinds of software while resolve is just resolve. But color management is a big subject, i found the tutorials by cullen kelly on youtube really helpful!

CSTs or RCM2 doesn't really matter. it's the same thing. one is global, the other one local.
ACES is great, but has some drawbacks.

For that many different sources I would recommend RCM2 for the following reasons:
ACES doesn't provide a few of the IDTs you will need with those cameras
CSTs make the pipeline flexible, but at the same time you have to make sure you set them up correctly on every single shot. copying the grade of another clip doesn't really work, because you're overwriting the CST settings of the current clip
Applying RCM2 globally on the other hand gives you a huge number of IDTs and it let's you copy grades and stuff without worrying about color spaces while you're grading. everything just works once you set up the IDTs in the beginning.

For the 8bit consumer camera sources, I'd recommend using the Cinematch OFX plugin to move those images to ARRI LogC3 AWG (and either use that as your working space, or then use CST to transform from there to your preferred working space).

RCM, ACES, and the Resolve CST — none of those have transforms designed for random 8bit consumer stuff like GoPro and Sony mirrorless. Even with the Sony 8bit stuff that is set to flavors of Slog and Sgamut, it's all pretty much being fudged by Sony. There's wide variation which the profiles in Cinematch take into account.

RCM and ACES I have no idea what the difference is other than them being so confusing

This may help.

Color space transform I see as a good, quick way to do things, but I don't know how accurate it is.

Using RCM is virtually the same thing if you apply the same pipeline manually with CSTs. One just runs in the background at various stages.

The idea behind proper color management for grading is not only to get all your individual material from the captured space to display space. It gives you a singular working space where you can grade in with consistent behaviour regardless of what source media it is applied on. In RCM or ACES this is project wide set up for you but as Jamie mentions these only have conversion options for the 'typical' scene referred gamut and gamma combinations. So your approach with CSTs in YRGB is totally the way to go as there's no alternative.

I don't know if you do this already but setting up the project properly matters even for manual management.
I would suggest choosing a scene referred grading space like DaVinci Wide Gamut/Intermediate or ARRI Wide Gamut/LogC3. Set your timeline working space to that and set your output space to whatever it is you're delivering, I assume Rec.709. This will ensure that the scopes read properly, you get the right tags on export and your color space aware grading tools function as they should.
Then the way you do this is up to you but I'd say group all your clips per camera and use the group pre-clip to convert from camera space to your working space. Then use the clip node tree to do grading and use the timeline node tree to convert from your working space to display. Prior to the display conversion on the timeline you can add any global look you'd want for the entire project. If you're also running titles and other elements that shouldn't receive grading this is a bit more cumbersome but I'd use the group post-clip for that part then. Only less ideal because it means the look needs to be replicated by the amount of groups you use.

Some things to note about CST:
- For your raw material pick something scene referred to decode to with the raw settings otherwise you lose dynamic range. REDWideGamutRGB / REDLog3G10 is probably the best choice there for RED.
- For all scene referred (log) spaces make sure forward or inverse OOTF is unchecked and tone mapping is off.
- For display referred conversions to the working space inverse OOTF needs to be checked. You'd use this for the clips that for example were recorded in Rec.709 with a curve that isn't available in the CST and will be graded manually to compensate.
- On the way out to Rec.709/Gamma 2.4 forward OOTF is on and you can use DaVinci tone mapping.

Hope that helps.

Another answer: color management was designed to avoid exactly the complications detailed above, which will not be well understood by beginning colorists.

For starters, try the defaults for rec. 709 output which come with Resolve Color Management. When your skills exceed those settings, some time in 2031, then look for greater customization. But until then, because there's so much else to learn, maybe keep it simple? Which is sort of the point of color management ... keeping it simple, so you're at a good starting point to grade log material, with a minimum of intervention. And your stuff, like so much amateur footage, won't look like incorrectly normalized log.

Another answer: color management was designed to avoid exactly the complications detailed above, which will not be well understood by beginning colorists.

For starters, try the defaults for rec. 709 output which come with Resolve Color Management. When your skills exceed those settings, some time in 2031, then look for greater customization. But until then, because there's so much else to learn, maybe keep it simple? Which is sort of the point of color management ... keeping it simple, so you're at a good starting point to grade log material, with a minimum of intervention. And your stuff, like so much amateur footage, won't look like incorrectly normalized log.

I think what you are saying is fair and valid. I would also maybe suggest going the other way and doing it all manually - ie no LUTs, no CMS, CSTs etc. Just start with whatever your camera(s) shot, viewing it in rec709 and manually use the tools to get something that looks nice and to your taste.

It's how I learnt ~10 years ago and while now all those other things help a lot (and to your point can do things like gamut and gamma mapping much more accurately) I think there is benefit in shaping the image completely by hand with curves and LGG etc.

I would now never not color manage in some way, but I think it can over complicate something technically when learning what the tools do and how to achieve what your taste desires in my opinion is more important and rewarding.
</my2c>

I think what you are saying is fair and valid. I would also maybe suggest going the other way and doing it all manually - ie no LUTs, no CMS, CSTs etc. Just start with whatever your camera(s) shot, viewing it in rec709 and manually use the tools to get something that looks nice and to your taste.

If the source footage is all baked in camera REC709, sure.
But if one is working from wide camera gamuts with imaginary primaries in log, especially if it's from different sources intercut, then some form of consistent profiled transform to a working color space and an ODT at the end of the chain should be used. With most of the wide gamut log formats, it is a fools errand to work by normalizing them for display using only LGG or curves. Plenty of practice using LGG and curves will happen anyway when grading in an intermediate color space under an ODT.

One thing i found lacking about RCM compared to a customized node setup -but i kinda hope i can be corrected on this one- is the ability to have a custom output conversion. For example, i often use this OFX plugin Look Designer, which has some different rec709 output conversions, or i can create/use a custom LUT, or even an ACES transform, or anything which gives me the desired look&feel. RCM just gives me one rec709 output flavour (with a few options), and i'm only able to alter this look before this conversion is happening in the pipeline, am i right?

One thing i found lacking about RCM compared to a customized node setup -but i kinda hope i can be corrected on this one- is the ability to have a custom output conversion. For example, i often use this OFX plugin Look Designer, which has some different rec709 output conversions, or i can create/use a custom LUT, or even an ACES transform, or anything which gives me the desired look&feel. RCM just gives me one rec709 output flavour (with a few options), and i'm only able to alter this look before this conversion is happening in the pipeline, am i right?

You could get around that by creating either an inverse transform at the end of the stream before your custom output nodes, or set the output color space to the same space as your timeline. This essentially gives you the benefit of auto conversions for source to timeline. I still prefer manual management personally though as it's more transparent and customizable.

Just start with whatever your camera(s) shot, viewing it in rec709 and manually use the tools to get something that looks nice and to your taste.

It's how I learnt ~10 years ago and while now all those other things help a lot (and to your point can do things like gamut and gamma mapping much more accurately) I think there is benefit in shaping the image completely by hand with curves and LGG etc.

I think that is the worst way to get into color grading. How your tools operate if you work on log in a display referred context are vastly different that proper management and in nearly all situations will get you into trouble when you want to achieve even the slightest of adjustments like exposure. On top of that the OP did say he's technically-minded so I'm assuming he isn't afraid to get his hands dirty and learn it properly.

So working in the intermediate color space with setting each camera in its properties with the right gamma and gamut is correct? It reminds me like how I edit raw photos in Camera raw where I edit in Prophoto RGB so I get the least color clipping but never export in it. Same idea here?

What I normally do currently is take in the log, use a CST to convert the color space only to R709 or R2020, then push the gamma around myself to taste so that I have a wide range of data to work with while getting correct color.

What differences should I expect output-wise if I switch to RCM? Seems like ACES won’t work for most of my super consumer cameras.

So working in the intermediate color space with setting each camera in its properties with the right gamma and gamut is correct? It reminds me like how I edit raw photos in Camera raw where I edit in Prophoto RGB so I get the least color clipping but never export in it. Same idea here?

You'd convert everything to the same intermediate space yes. It is to an extent comparable to photo editing yea.

What I normally do currently is take in the log, use a CST to convert the color space only to R709 or R2020, then push the gamma around myself to taste so that I have a wide range of data to work with while getting correct color.

This is the same as grading display referred which I would not recommend to get used to like I mentioned in my previous post. You want to rely on the tone mapping the CST or RCM/ACES gives you to properly map/distribute the captured dynamic range into the limited display space. If you incorporate this into your grading itself it becomes very tricky and unpredictable to push the image around. If you let colormanagement handle this, for log footage using your offset wheel before you apply the output CST will act in the same way as changing exposure and feels very natural and intuitive.

What differences should I expect output-wise if I switch to RCM? Seems like ACES won’t work for most of my super consumer cameras.

None. As I mentioned in my first post, RCM is essentially the same as using CSTs but the conversions happen in the background. RCM will only complicate things for you because most of your material is shot in gamut+gamma combinations that RCM doesn't support.

If you're interested in understanding what RCM (DaVinci DRT) or ACES are doing I can recommend watching this talk. Most videos of "youtubers" either explain it way to dumbed down or they are full of technically false statements.

Thanks. I’ll check out the video!