Is there any hope for flame in the future for compositing?

You have to attach them in the media pool, but Resolve can see multi-layer EXR files and address them individually in the node tree.

I suggest to back up to the 10,000ft level. What are we trying to prove here?

As long as enough people keep using Flame for ADSK to keep paying the devs to maintain and improve it, there will be Flame. Itā€™s not going anywhere soon. For one, because it has some things certain teams just need.

The same can be said about Nuke. Itā€™s being used by enough people that the Foundry will keep making it.

So then the question just is - which one do you want to use and become fluent in. Depends on your work and your style. You can do one, the other, or both. Itā€™s really that simple.

I mean itā€™s a fun discussion from time to time, but not sure itā€™s going to lead anywhere meaningful.

5 Likes

Agree. I do think is meaningful to expose artists to useful tools/features, otherwise unknown to them.

1 Like

Can you point us to a good example? I searched Google and YouTube and only came about one reference for EXR / Nuke sub channels as the ability to add alpha mattes to various AOVs. But not sure thatā€™s what you are referring to. Doesnā€™t seem to have much coverage anywhere. Unless Iā€™m not searching for the proper terminology.

So, I guess you are taking about EXRs. Is that what you mean by ā€œcan be used in other softwareā€
Or are you talking about Nukeā€™s handling of multi-channel layers in a single stream?
This is getting confusing now.

Mattes:

Nuke vs Flame:
In the end, a Nuke artist with a good pipeline and knowledge, of using multichannels in a productive way, will beat Flame. But also a Flame artist with a good pipeline and knowledge, to keep the setup clean with only 3 channels per link, will beat Nuke.

I have to admit that itā€™s frustrating to only have a front and matte in action. It would be amazing to be able to transform more than RGBA without jumping through hoops.

3 Likes

OK. I think this is not a specific feature per-se, but more the fact that Nuke is very transparent with itā€™s channel layout and makes it easy to access and manipulate them. More so than other apps.

They divide each connection between nodes into layers and channels within layers. Generally the node defines the inputs and outputs it understands, but beyond that with the shuffle node you have extensive flexibility to access and remap them, and as does the viewer give you full access.

I believe you can even add your own custom layers and channels and then fill them with the shuffle node from other sources to create your own hybrid data sets.

That is very different than Flame or Resolve which tries to overlay a fair amount of semantics on the connections and babysit you in terms of what can be connected to what. Resolve wonā€™t let you mix color and matte channels at will in some cases. Flame makes it much more onerous to split a node connection, and you need context specific nodes for that.

That does make Nuke easier to work with in very large comp scenarios, as youā€™re wasting less real estate in convincing it what needs to be done. Same with the fact that in Nuke color and matte can travel in one pipe rather than requiring two is a useful declutter.

All this is to say, I donā€™t Nuke has any special sauce on EXRs. It just makes it easier to work with node connections and how theyā€™re structured (names of channels and sub-channels vs. layers and channels aside).

Great explanation. Nuke layer system allow almost any tool that deals with channels to modify the layer structure. That means the a Grade node for example can be used to create, delete or modify layers.
Nuke tries follows the EXR sepcification fairly close, although I think some features are still not supported. Maybe what @maz friend is referring to is the ability for Nuke to deal with Multipart EXRs. In Nuke they are called Views. Multipart EXRs can contain several images streams, each with its own set of layers and channels. Each stream can even have separate bit depths and Data Windows. Not sure if layers can have different bit depth.

The most common use of Multipart is Stereo work. Nuke will treat each eye stream as a separate View. Nuke Multipart structure can also be used in VR and VP work when doing stitching of camera angles for spherical projections. Each camera angle is structured as a View in addition to the Main one. This way each angle along with the camera and its metadata can be passed down the stream for further work.

Here are some multipart EXR examples:
https://openexr.com/en/latest/_test_images/v2/Stereo/composited.html

2 Likes

Interesting. Hadnā€™t come across those before. Would there be any use case for multi-part EXRs (i.e. multiple streams) outside of stereo and VR?

I guess maybe you could combine original and graded versions of the stream? Or HDR and SDR grades? Not sure this would be useful. Makes for massive files and fewer tools that can open hand handle them.

Something like that will come eventuallyā€¦ just like it works for audio. I am expecting an MXF like container to drop for EXR in the not so distant future.

1 Like

Iā€™ve seen the views used for different versions of a shot. In a commercial you can have new and non-new views. Itā€™s very flexible and useful.

I would love it if a layer could pass through Action and carry more than just RGBA.

1 Like

Maybe. But that green screen comp is pretty poor.

So hereā€™s the thing: at the mill if there was an impossible shot to do, it would get done in flame. I think the image integrator subtitle needs to be brought back.

I love the fact of single pipe in nuke. It really makes copying and pasting bits of a batch easy. I love the ability in nuke to copy and paste a script into an email. I am starting to warm to the idea of the bounding box too. But overall, for me, using the nuke still feels a bit like a flow diagram following someone elseā€™s instructions.

With flame, frustrating as it can be, it feels like you can anything work. Though you do need quite a lot of tacit knowledge.

Whatā€™s great about this forum is that we are sharing that tacit knowledge. What we might need is some keen library students to tag the shit out of it and make it a bit more easy to find things for newer users.

And we probably need another anniversary version of flame to iron out all the idiosyncrasies.

2 Likes

Jimmy Fallon What GIF by The Tonight Show Starring Jimmy Fallon

2 Likes

I applaud your efforts @randy. Whilst this does serve a similar purpose, my thoughts were geared toward the open source nature of the knowledge we have all shared.

1 Like

Iā€™m sorryā€¦Iā€™m not following.

Logik pro costs money. This forum does not (although a lot of us donate).

Also the information already exists here in the forum but as an archive it is sometimes difficult to find. Hence my desire, albeit realistically it is probably difficult to achieve, to have it given a thorough indexing.

3 Likes

Good thing is, asking something on the forum or researching it after some time later wonā€™t be blocked by a paywall from logik pro and will be answered by the community.

Therefore the pro concept for learning flame and the logik community for working with flame isnā€™t that bad afterall that way. (personal opinion)

2 Likes

Iā€™m in a discord forum for Warp Fusion, an AI character replacement technique. They have an AI bot that you can ask a question, and it will scan the forum, give you and insanely coherent precise answer, and highlight to 4 most pertinent posts/threads for you.

Check it out:

Can we get something like that working here?

4 Likes

Unfortunately, the short answer is no.

Iā€™ve looked into various bots and opening up the forums as a data set to specifically the Discourse enterprise plan. The challenge is that alone is between $2000 and $5000 per month, minimum.

With the amount of data that is already existing, the amount of monthly users, the amount of page views, and the fact that the forms are free makes that cost prohibitive.

They are, of course, smaller and more targeted ways of attacking this problem, but I havenā€™t found any of that would really turn the needles. Especially when the search on the site is super good.

I mean, the ultimate question is, what canā€™t you achieve with our given tools right now? Itā€™s far more helpful to know what your challenges are.

1 Like