Wednesday, 14 November 2007

Day313 AEAF Day2

Australian Effects and Animation Festival - Wed
Another great day at the event. This time the Mac/PC split of the presenters was more even, 3 MacbookPros, 1 Dell, 1 IBM and one desktop PC (which only ran html btw). Another thing of note is the dominance of the Guys, all the presenters have been male and the audience is mostly guys as well. I still find this most perplexing, given the story telling nature - oh well.

Every one of the presenters ran out of time and there was only time for a very short burst of Q&A from the floor. I wish there was a way to give each of them longer actually, anyhow onto the breakdown of what I got from each of the sessions.

-

Theo Bialak : Sony Pictures Imageworks
This was always going to be a great session and Theo didnt disappoint in showing off the splendour that is BEOWULF. First up he talked through the epic mocap work they had to do in generating the data used to drive all the characters. The mocap set was big by these standards and sported between 224 and 260 cameras capturing the body and facial data of the shots. Some shots has 30+ characters on the set at a time, which is why so many cameras were required. They also had special new tech to detect the eye movements and blinks from electrical impulses. The setup had each actor sporting 250 markers and was generating some 350GB of daya each day. This added up to 17.4TB of mocap data alone for the project.

The modeling (in Maya) required very high-res and detailed models with procedural elements for things like chainmail. Theo walked us through several characters and details as they evolved from concept to final shot. They used a system of parts which could be combined to build the array of extras and minor characters everywhere. They built a new custom fire renderer and again built a library of fire elements for many of the shots and uses.

Theo covered briefly the second team working on the 3D (think IMAX) version over the 2.5 years of production. He also mentioned in Q&A how they tried to keep the number of passes down to a minimum to keep a handle on the scale/scope. They had some 6500 processors working on rendering - YIKES and even had 60+ lighting artists working at once on shots which just shows the scale.


[still images of Beowulf from the Rotten Tomatoes site]

The stills do look stunning, but for me the animation is butting heads pretty hard with the Uncanny Valley in a big way. This was covered only very briefly in Q&A, but only fleetingly - I would have loved more. The animation still feels solidly CG to me and thus feels awkward in much the same way as this film's only real peers Final Fantasy and The Polar Express.

What remains now is to charge off and see the film!!!

-

Allan McKay : Catastrophic FX
This session was a change of pace with Allan walking us through some of the effects that can be created with FumeFX (he was using 3DsMax). He showed us clever work in both particles and fluids, how the simulating is the slow part. He ran through plenty of fire and smoke/dust examples plus a bunch with particles like Superman Returns' supernova shot with its 700 million particles :-)

-

Damian Allen
This session was more of a show'n'tell for Apple's Motion3 software that ships as part of the Final Cut Studio. This compositor uses the graphics card firepower to show us in realtime what we are getting and can do some fun and intuitive things with behaviours and library elements in 2D and 3D.

-

Evan Jacobs
Back into cinema-land with Resident Evil: Extinction and the effects work required to bring the latest in the series to us. Evan gave us a fun tour through some of the CG challenges and how he faught to keep as many real elements in shots as possible, be they actors or models. He felt that this delivered better results than fully CG shots.

There were lots of anecdotes around the desert shooting in Mexico, though the clear blue skies did work nicely for them as a built-in blue screen set. They seemed quite precious about Milla and seemed to face replace body doubles when the shooting was too unpleasant. They used Massive for their CG zombies and there are lots of them.

He did mention that at one point they had to scan 22,000 frames of footage for a scene in CG post which at $1 a frame adds up fast. In fact Evan thought that alot more was being left to post now rather than plan things.

-

Sebastian Sylwan : AutoDesk
Sebastian gave us an interesting session with a different feel, coming from AutoDesk, but with his attention on the future and industry trends. He covered ideas around the concept of pipelines and innovation in the industry. He had examples like digital end-to-end, HDRI OpenEXR, Gigapixel Imaging etc. His view was the the increasing reliance on post wasnt simply lack of planning - but rather because so much more creative control can be exercised at this time. He did poke his talk into Uncanny Valley land again (this was the theme of the day really).

Check out the fun of Microsoft's xRez beta - much like Google Earth in logic - very cool.

-

Steven Stahlberg
This was one fo the sessions I have been looking forward to the most for what seems like months now. Steven's work is simply stunning and everyone could learn from him. His talk began with some tips on how to stay fresh, alive and learning - 'fertilize your mind'.

Steven covered the Uncanny Valley again as well and coverd some very interesting ideas around archetypes and the anatomy of the Male, Female and Baby.

As far as modelling goes, one tip was to use a perspective camera rather than the default orthographic views (say in Maya) as the reference images would have been taken with a real perspective camera. He covered in great detail from his course notes issues around modelling the female form (with all to frequent apologies for what could be seen as sexist observations :-)
He likes the more organic feel of tris rather than quads. He uses a fold based topology for faces and talked about plenty of the errors his students make. It is not just his new students who miss things though, he gave us examples of things like the bumpy edge of the ribcage. His anotomy lessons on things like the 'carrying angle' where girls have a different angle for their cubital fussa - ie their elbows face up or out more than guys.

Steven gave us plenty of sites to explore such as:
3d.sk : for all that reference imagery
character design blog : Randall Sly's coool blog
beauty check : research on facial attractiveness
face research : more on facial beauty
beauty analysis : Marquardt Beauty Analysis
rey bustos : anatomy education
plus
android blues : Steven's site


[some images of Steven's from his android blues site]

Steven did have a little fun with the golden mean advocates, showing reference to studies showing that we dont really put much store in the ratio - but actually favour others which is interesting.

I soooo wish we could have had him talk for longer - he had us all intrigued and had a heap more to reveal on his techniques and observations on anatomy and generating the form in 2D and 3D.

So this session was a great way to finish day2 - drool.

---

My collectors edition of Crysis is sitting at EB Games waiting for me - I will be grabbing it tomorrow, I cant wait to both see it in action and to play with the creation tools.

I also read a bunch more of my Machinima book - it is quite informative and a great read. I feel that the depth of thought required to build a narrative in this (and many other forms) is higher than I had really given credit to before - looks awesome though. I liked that none of the engines are really suitable - they all have good and bad points - it is part of the Machinima lore really...

No comments: