I’m inclined to support any of these measures, but I’ve found deep renders to be exceptionally rare. When I asked about them at the Mill, the CG guys said the files were huge and I got the impression that they weren’t useful.
I’m sure that in some studios they are both common and useful, but they aren’t in my orbit, so I lean towards Chris’s USD goals simply for the side benefits that might come from an overhaul of Action.
Is it fair to say that ‘deep’ has it’s place in certain genres of CG and is definitely up on the sophistication scale. So if you work in that type of genre, like @milanesa does, then they’re daily diet, but there will probably be many CG artists who can have an entire career without ever encountering deep.
So in one way we’d be solving for the 1%, but then Flame is also at home in the 1%, just may be not the same 1%…
And may I suggest we make some popcorn and see how Resolve/Fusion fairs with deep before we join the party. I have no doubt that ADSK would do a better job, but sometimes it’s still good not to be the first out of the trench. Let someone else volunteer to find the landmines. And they seem so eager to volunteer 
Honestly, I couldn’t care which order they were done in as I think that both are anctually important if Flame is to compete in a heavy compositing environment. If Deep came a couple of years later than USD then so be it. As long as in this improved 3D workspace there is a way to unwrap UVs, like you can do in Nuke & Fusion. Yes, I know there is a workaround but it’s not quite the same thing.
Deep comps can be expensive in that these files are big in size. It would be also great if we could projection
Paint on a model, sort of like you could do with the Mari/nuke connectIon with proper UV’s of course. So many times I could use a feature like that. Freeze a painted frame at one point in a shot, bake that onto a model and find your next clean frame to bake ontop of that and paint away. VDB support is a must have. Yes it would be great to have EXR’s that we could pass all channel info downstream from node to node like nuke.
Right, I think what everyone is saying is that the current Action toolkit is in desperate need of a modernization and improve things like UV projection without crazy setup, etc. Whether along the way we pick up deep or USD is kind of a cherry on the cake.
With @FrancisBouillon back, the timing may be right.
Well said @allklier
My concern is this, I know a few ex-Flame Finishing Editors who now work in Resolve. At first they may have just jumped in Resolve to Grade. Then they started jumping into Resolve for the odd codec that Flame couldn’t read. Then they jumped into Resolve because it conformed faster. Then they jumped into Resolve as they needed to master an IMF. Then they jumped into Resolve because they needed to listen to Dolby Atmos. Until one day, they opened Resolve and decided there was no point in leaving it anymore. Essentially what was required kept evolving and the Flame toolset didn’t keep up.
So now we are seeing both Nuke & Fusion with compositing toolsets that people may just jump onto just to…
I love Flame. I have no qualms with the dev team at all, have always expressed my gratitude and support for them as they listen and they implement what they can. When they add a tool it is mostly well thought out and well integrated. However, it’s a real shame that we so often see on Logik people saying dev time and money is limited so they would prefer the dev team to focus on only one of x, y or Z but then you see all the tools you are asking for released in other products ported to all 3 major operating systems. It would be nice to be discussing all the crazy unique new features Flame has introduced.
So I would like Flame to have both USD and Deep as well as some other cool tools that nothing else has (e.g. some cool photogrammetry and/or NERF + Gaussian splatting utilising crazy ass tech that you can feed a moving shot into and it creates 3D scene within Action that you can do amazing shit with) and I’m not going to apologise for suggesting them or settle for having to choose just one when others get the whole set.
I totally hear you.
I have extremely mixed feelings on BMD for it’s negative impact on the industry. And I held out as long as I could preferring to work in Avid. But in the past few weeks I shut off Avid software and hardware, and am now relying on Resolve to edit, rather begrudingly. But for exactly the logic you mention. I have 3 Resolve license, had them for a long time, and have used Resolve in the past for grading (before Mistika, and Flame, and Baselight). But at the moment Resolve offers the best feature set for editing with some remote editors that meets my business needs better than any other solution out there. So I’m grinding my teeth while launching Resolve.
The way I like to work with Deep in Nuke comp is to treat it like another AOV alongside a standard multichannel beauty render. For non-volumetric renders, we’d render a sidecar Deep file with only the Deep data and b&w color data similar to this technique: Deepexr and exr: the easy way
Your Deep files are much smaller/manageable this way since you only carry 1 sample for color data.
When you do need to take advantage of a Deep toolset (for bokeh, generating holdouts, fog, complex layering, etc. ) you can promote your standard rgba+ comp to Deep with a DeepRecolor node and your b&w sidecar Deep file. You only have 1 sample of color data but that is ok for the vast majority of comp needs.
For generating holdouts, I’d usually do that in a parallel Deep stream and render that to single channel exr files after a DeepToImage node. Or bake those mattes into a custom cryptomatte pass for later use.
You can get away with not rerendering the Deep AOV again if anim and camera is unchanged (as you do with DATA_AOV passes)
I guess many are under the assumption Deeps are super heavy. They can be, but they can also be small and easy to work with, like any image file. The workflow @mybikeislost describe is common and many times necessary. DeepID is a good example of small file footprint. Way smaller than Cryptomatte and more versatile.
Regarding USD, I was referring to a more native and technical implementation of it; hence my initial question. The consensus from the discussion is to have a better 3d environment/toolset in Flame, which I agree comes before Deep if I need to choose.