There were a few talks on the VFX involved in creating Avatar…
Presented by Shawn Dunn. This covered a broad range of the work done on the film – from the motion capture (or performance capture if your feeling insecure) right up to rendering. Most of the stuff talked about here has been covered already in articles by Cinefex and podcasts with fxguide. The really interesting stuff was the work-in-progress shots which showed how things progressed.
Presented by Paul Kavanagh (Animation Supervisor). This covered the work done by ILM on the film.
How the work was divided up between Weta and ILM? Based on what shots ILM could do rather than specific sequences. Anything that didn’t feature the Navi/Avatars was a potential shot for ILM. The majority of work was in the final battle sequence, but they did work on other shots like the arrival of the shuttle to Hellsgate. The only Navi shots they handled were when the Navi were really small in frame.
Developing flight cycles for Banshee. Originally created controller on rig which allowed an animator to automatically blend between cycles – transitions looked too linear so they manually created the transition and not just one transition between two cycles – for example in transitioning between a glide and a dive, there would need to be a transition which went from a glide-to-dive and another transition which went from dive-to-glide – there were nine different flight cycles which required a total of 72 transitions. They used the Trax Editor in Maya – “a horrible tool… any Autodesk representatives in the room, could you please make it better?”
Tracer fire. Showed some of the reference they used for the tracer fire… Helicopter Tracer Fire, Mini Gun Tracer Fire, Nighttime Tracer Fire and Russian Heavy Machine Gun.
CG Pyro. Tried at first using the CG explosions done for Star Trek – which didn’t look very good. Didn’t start working until Chris Horvarth had a shot at it and they started getting the fluid interaction between the explosion, the amp-suit and interior of the ship.
The Walnut Incident. Due to CG pyro still not quite working they didn’t have any FX TDs to work on the downwash effect at the start of the film. So John Knoll set up a test shoot to film some practical elements and find something that would create the same effect. When it was mentioned to James Cameron that this was what ILM were doing, he had a few choice words on say on the matter – along the lines of “That isn’t going to work” but using much more ‘bleeping’ and asked John Knoll to show the results of the element shoot the very next day. The next day John compiled the downwash elements together to show how he planned to use the elements in the composite. When Jim saw this he calmed down “That looks quite good”, “It scales really nicely”, “What did you use?” John Knoll: “Walnuts. They’re crushed walnuts.”
Funnies. Paul showed a few of the silly submissions that were done by ILM, hopefully these make it onto the DVD special editions – stuff like Facebook conversations and Twitter feeds for the characters appearing on the monitors in the film, In ‘N Out Burger signs in the middle of the jungle, outtakes from the mocap sessions where one of the soldiers briefly breaks into a little dance and amp-suits running around with guns up their bottoms.
Presented by Shawn Dunn (Head of Layout and Animation Technologies). I had worked occasionally with Shawn at Weta Digital when he was Lead Animation Technical Director – his role was to create tools for the animators (retime animation, flip poses, etc) and to develop the animation pipeline. The times that I had worked with Shawn was during the occasions when assets from Models went straight to Animation rather than through Creatures – for example in order to speed up the previs process Models created logical pivot points rather than freeze the pivots at the origin – this would allow the Previs animators to select a piece of geometry and automatically create an animation control based on that pivot point.
I left Weta about the time the Layout department was just getting started, so I already had a fair idea of what it’s role was intended to be and I was interested to see how it had changed over the last two years. This is the shot Shawn used to demonstrate the Layout pipeline.
Templates. Once James Cameron had shot the performances and created his cameras Lightstorm (LS) sent a template file to Weta via FTP which contained the set, the motion capture data and the camera for that shot plus a quicktime to show what had been recorded by Lightstorm and what Jim wanted. Weta took these template files and used them as the starting point for rest of the Weta pipeline.
Check for mistakes. The first thing Weta had to do was take the template and do a playblast – this was to visually check that Lightstorm had sent everything and nothing was incorrect or missing from the template.
Start working. Once everything checked out, the Layout department did a playblast which showed which generic assets (plants) used by Lightstorm were missing from Weta’s asset library, these assets would be marked down and given to PreProduction (Models, Textures and Shaders) who would start building them. Often Lightstorm would randomly scale assets to dress scenes (think a tiny scrub scaled up into a huge tree), this type of scaling wouldn’t work for a hero asset, so variants needed to be created which would work for that scale. As well as generic assets, Models would also be given scene specific geometry to start work on, either specific sets that needed to be built or the modelling of terrain (ground, branches, etc) which the characters stepped on. The terrain modelling was done to the animation, rather than having the animation done according to the geometry – the mid-res terrain was sent back to Layout as well as passed on to Textures and Shaders.
Giving a shot the third dimension. During this Layout would start layering in the rest of scene to give it depth as Lightstorm only dressed the immediate surroundings of the characters, James Cameron was very particular about seeing into the far distance on shots. Once done Layout would publish an XML scene which was given to Lighting, this contained all the various components required to render the environment and animation. The XML file just described the placement of all the objects in the scene as bounding boxes and referred to rib archives which were dynamically loaded and unloaded at rendertime.
Keep in mind that this pipeline from start to finish is fairly non-linear.
The following diagram is one I created to highlight the breakdown of a shot based on what was said.
Scene Specific. In this shot the branch that the characters run on needed to be modelled. This model was quite hires and sculpted in Maya and Mudbox.
Asset Exists. These plants already existed in the Asset library and could be rendered straight away.
Asset Doesn’t Exist. These plants had yet to be created and were sent to PreProduction.
Additional Dressing required. Layout had to take existing plant assets and start layering them in to the shot.
Using Matte Paintings. Although a lot of what was rendered did exist in 3D, there were times where they used Digital Matte Paintings when parallax wasn’t an issue. In order to help the matte painters, they would render out assets with 8 different keylight directions, these elements could be freely used by the matte painters to fill in shots.
This talk by John Bruno (Visual Effects Supervisor) should have probably been called “Random Anecdotes about Avatar” as he didn’t really go into Weta’s “way” of doing things. Not that it wasn’t interesting, just a little un-focused.
So what did he talk about…
Scale differences. At the end of the film when Ney’tiri picks up Jake she originally looked the same size as him – even though she’s suppose to be 9 feet tall. And then in the next shot her hand looked unnaturally large compared to Jake’s – oh, there was a lesson learnt here, the lesson is that it’s “not that easy to just scale the hand of the character rig”, instead the solution was to adjust her hand so the largeness of it was less distracting.
Warpaint. The original warpaint design of Ney’tiri wasn’t very flattering – it made her nose look too wide. Gino Acevedo came up with a design which worked and is the one seen in the movie.
The Walnut Incidence. Yes. The same anecdote given by Paul Kavanagh in the ILM talk. Maybe some people didn’t attend the ILM talk.
The Usual Suspects. Once the film is first edited, Jim Cameron invites a select group of people to see the film. This screening is to get brutally honest feedback about the film, Jim only invites people he knows will still be friendly with him after any arguments that result from this screening.