The short short version (’cause there was too much to process)
The CGA Conference has been a blast – great organization, smooth and cool runnings (pun intended) and outstanding talks.
This has been an artist conference, by the artists, for the artists in the technologically advanced side of business we are in. Practical solutions and practical applications were front and center, with a “this is how you do it to get the best results” approach across the board. The VFX peeps are a pragmatic bunch aren’t they?
I met and heard some brilliant people from all around the world, and my rolodex of highly skilled artists and entire facilities is now at least twice as big. Lot of potential partnership for an array of tasks are now in play, and shooting in Serbia now has a lot added perks in high-tech support services from mocap to LIDAR (which apparently existed – unbeknownst to us all – since 2003 in Serbia – and we didn’t have a clue!).
VFX supervision talk
Yep, you guessed it. It was me. I went through this old business stuff to make the process of bidding, supervising and finishing a job a bit more transparent with a brief introduction of the Autodesk Flame workflow, my software of choice. I went with the industry standards of today, and rolled with it. Turns out – it was spoken about as one of the most important things to know, because it ended up being the only talk about the “business” side of our business, client relationships, why things are done the way they’re done, etc. I did similar talks before, but this one I feel I’ve done a bit of a better job then usual. I’m already sweating from seeing the video recording if there is one. Finger’s crossed it’s at least ok, as I still struggle with stage issues.
Bogdan Mihajlović of qlbeans.com, the motion tracking studio based in Belgrade, Serbia had a really immersive talk on how can we in the VFX supervision business make his and our lives easier. There were several showcase studies of very different challenges he’s facing when doing a job. I know like 10x more then I knew before about how to do a better job now. Top, as the Dutch would say! FX Guide published an article on the talk I have personally very much enjoyed, go over there and check it out.
Also many talks on the current workflows in indie cinema, how to become a better artist (hint – work, work, work) and many more great topics were addressed in only two days. It was intense, but somehow it felt like we needed more.
The strainer and the silver car
First there was an image of your kitchen metal strainer over a similarly colored car. Then, there was an image of the resulting matte of that same strainer extracted to a level a human would have a huge problem achieving. Metal over silver, like a key pulled off a green screen. Think about it, we’re talking telling the AI to give you mattes of all the humans in the shot, letting it work on it for a while, and the using those mattes in your comps. Level of detail was outstanding, speed was unimpressive as the best boxes we currently have spend days to figure out what’s the strainer and what’s the car. Then I saw a few others with a woman with a lot of hair, and then another one, and another one, and another one. It’s probably still years away, not months, but we’re talking a few more years and we never have a conversation with our clients about how using roto adds a lot of overhead. Frikkin’ AI will extract it for ya 🙂 Then it’s all rendering hours we have to bid.
I am an actual beta tester of this technology now, so I’ll be able to reveal more soon enough.
The People of the Deep (Learning and AI) World
Now, that was the practical part, something we can apply today without going to jail or having an ethical discussion. However the face-replacement stuff was a whole different beast. Mike Seymour (fxguide.com), Vladimir Mastilović (3Lateral) and Ben Lumsden (Epic Games) had a great talk on the state of machine computing and deep learning with some incredible points on both the technical and moral segment of this exciting new technology field. The Q & A was particularly interesting as most of the audience questions were about the moral and ethical issues of face replacement applications, and ways to prevent abuse, I waltzed in with my American fear-mongering mentioning Stephen Hawking’s prediction that “AI is the worst thing that will happen to human kind” (paraphrasing here, sorry if it’s not 100% accurate) and his idea that we need to establish a set of guidelines that will help us make sure those AI’s don’t kill us all in some Matrix-like scenario. Didn’t really get a reply there, but I did get *a* reply. And it was mostly about “how we’re not there yet, and there’s nothing to worry about – yet”.
Robots with AI were mentioned, and 3Lateral’s pet project of creating the 1000 archetypes of humans which I can’t explain in detail, you should hit them up for more info. Cool stuff though. Really immersive talk, really hard to re-tell.
Unreal Engine guys appeared with an obvious (if carefully NOT mentioned) roadmap to RTX. So did the guys behind Arnold, and the guys behind VRay. The future is real-time, very obviously. Lighters, realists, everyone and anyone with a know-how in this field including myself was shaking form excitement.
From my beginnings back in the 90s with the demo-scene – to this – the prophecy is coming to life during our lives.
Real-time is officially everywhere.