Interview with "The Matrix" VFX supervisor John Gaeta

[azazel]

Great reading, thanks for the link:thumbsup:

[BrandonD]

Hehe, this is the same guy that opened his commentary on "The Matrix" with "I'm John Gaeta and I did all of the effects in this film." So much for being humble or respecting the hard work of the artists that helped him get that golden statue. Oh well.

[malducin]

Not only the DVD but also his acceptance speech for the Academy Award, which was ... how to say it tactfully? was "interesting". Not to mention his other rather strange comments he has said over the years.

[RobertoOrtiz]

...or the presentation he gave at the AWA meeting in SIGGRAPH 99.

It was very “unique”....

[floguyep]

That was an IN-CREDIBLE article!! So interesting!

[beaker]

Originally posted by RobertoOrtiz
...or the presentation he gave at the AWA meeting in SIGGRAPH 99.

It was very “unique”....
I remember the one he gave at the Softimage UG meeting that year at siggraph(Still my favorite softimage meeting to this date, all the ones after that have been crap and boring). Walking around in that flight suit(or was it a nascar jumpsuit?) all over the place. He talks extremely confident and knows what he is talking about(and definatly flaunts it). He reminds me of a Richard Belzer, fast talking and such.

[RobertoOrtiz]

Was it the Softimage meeting?
My God that Siggraph is a complete blur to me.

[Seven]

Thats an amazing article, but this part is really interesting:

'The standard way of simulating the world in CG is to build it from the inside out, by assembling forms out of polygons and applying computer-simulated textures and lighting. The ESC team took a radically different path, loading as much of the real world as possible into the computer first, building from the outside in. This approach, known as image-based rendering'

Now that doesnt make much sense to me, they havent explained how they do it.

Does anyone get what they are on about ?

Sev

[RobertoOrtiz]

Please someone correct me if I am wrong but it sounds like a fancy way of explaining High Dynamyc Range Images (HDRI ) to explain the ilumination model and modelling done using Lidar based techniques and a heavy dose of Photogrammetry.

Right?
:rolleyes:

But these methods can not be what they are talking about, since they have been used for a while now.

-Roberto

[BrandonD]

Actually, Image Based Rendering is the stuff that BUF pioneered on the Fincher films. So yes a bit of photogrammetry.

[Lunatique]

Damn, you beat me to it. Good thing I checked to see if someone already posted it.

[malducin]

It was probably the Softimage one since I think I went to at least part of AWGUA and I don't remember Gaeta on the lineup but I could be wrong. Seriously, that guy needs to hire a diplomat or public relations agent at least.

As far as those parts of the article they might be a bit misleading and the technology has been explained in stuff like the DVD and Cinefex. First of all image based techniques is just a generic term, just like global illumination. You can do GI with a variety of methods: radiosity, hybrid renderers, photon mapping etc. Just the same image based rendering (or even image based techniques) is a generic term that encompasses several things.

HDRI is usually equated with image based rendering but that is not so. HDRI is just a format to capture a bigger dynamic range of lighting or color that traditional image files. HDRI is usually used in image based rendering but one doesn't presupose the other.

Photogrametry is in basic terms a way of making measurements from a set of photos. You can say it's an image based modeling technique. It's used a lot in cartography and the like. In some science museums you can see 2 photos taken by a plane (from two separate cameras) and then look through some glasses and see a 3D pic of the terrain. More or less in the same way measurements can be made of terrains, hence you could use it to measure environments and sets.

Now image based rendering is basicly rendering the image using thwe HDRI image instead of using any lights. You "wrap" the texture around your environment and then sample the hemisphere to see what the illumination is at that point. It works because in essence the HDRI image represents all the light (either direct or reflected) around your subjects.

Now the article goes a bit extreme and misleading since it's not like ESC is the first or the innovator. Probably the Matrix (by Manex) was the first large scale use of image based techniques in a film but certainly not the only one. The X-Men Senator death sequence by Digital Domain was image based rendered, Centropolis did muchthe same thing for Eight Legged Freaks, ILM in several sequences of Jurassic Park 3 and Pearl Harbor and there are other examples. As BrandonD, BUF has done it quite a bit in Panic Room and Fight Club.
As far as image based modeling, Disney's Dinosaur and Imageworks in Spider_man used photogrammetry and End of Days used Lidar for the ending sequence in the Church.

[aurora]

I went to the book store last night and checked out the May issue of 'Wired'. The article in there was longer and went into ALOT more detail on the Photogrammetry and the creation of the Burley Brawl. It also stated that they rendered with Mental Ray, interesting, very interesting.
Tonight (Thurs 10) on ET Tonight they will be broadcasting the premier of the new trailer which they hinted will have more coverage of the Burley Brawl in it. Man this is going to be a long day waiting for that!!!

[BrandonD]

Actually it makes complete sense to use Mental Ray for Image Based Rendering. Again BUF pioneered it's use in the Fincher films, so it's totally conceivable that ESC said "hey, what did they use and how did they do it?"

[aurora]

What made it interesting to me is the fact that I'm a LW user alot more then I'm a Maya user which now has Mental Ray. After seeing the feature list for Maya5 and (NO FLAME WARS PLEASE) not seeing LW8's yet. Well I have no desire to start another one of those threads so I just left it at 'interesting'.

[malducin]

Well I don't think they asked themselves how did they do it. Mainly because one of the core R&D guys at Manex was one of the researchers with Paul Devebec, who had worked on some of the image based research at Berkely. The Campanille movie was presented in SIGGRAPH 97. Borushkov (sp?) then joined Manex to work on that aspect for the film.

After the Manex fiasco I remember there was a dispute that Manex said they couldn't take any of the technology to ESC. I don't know how it ended but probably the ESC guys had to build a lot of things from scratch. They probably saw what could be bought off shelf, in the case of the rendering see what package could be used for image based rendering.

[RobertoOrtiz]

Aurora it is standard policy from NewTek to never release a feature list from upcoming versions.

This has been the case for ALL versions since version 1.0.

And going back to the subject


I wonder what TRULLY new technology will be presented on the new Matrix movie. Or are they going to refine older techniques?


-Roberto

[aurora]

Hey , I'm not hacking into Newtek, as a Software company (where I'm a programmer) we NEVER give a feature update either so I totally understand and side with Newtek. Thats also why I tried to state that I did not want to start another one of 'those' threads up.

As for the Matrix from what I gathered from the article in Wired Mag itself it sounds like that took alot of existing tech and greatly expaned on it to whole new planes of functionality. It does not sound like the tech was really all that new. Just what they accomplished as a whole was.

[CGTalk Moderation]

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.