Simply Maya User Community - Narnia 2: The Insider View

Hi Guys

I know I don't post much here anymore, work is normally a fairly flat out affair :)

I just wanted to share with you what I got up to on Narnia 2 and give you a bit of an insight into what we did on the film.

On the show I was put in charge of creating the CG Army that appears on the riverbanks towards the end of the movie (The bit with the watergod). We were initially supposed to also do the guys that get thrown around on the bridge but those were eventually done by the guys at Scanline.

The crowds were done with Massive and we started off literally with nothing: no pipeline, no models, no rigs, no mocap etc. etc.

So the first step for me was to take the idea and start to flesh out exactly what was needed for a pipeline. We did a quick test shoot at the audiomotion mocap studio in order to get some test data. From this I built a quick agent in massive that would walk around some terrain and do simple stuff. It was purely for a test so it was very rough and ready.

Once the test agent was done I set about designing the pipeline, I was working alone at this point and there was tons to be done! The way that data is spat out of massive is fairly simple after you run it you end up with some rib files and some animation files which describe each agent's motion for the duration of the sim/frame. What was important here was to get that data out of the Massive pipeline and into our normal renderman pipeline, we used Massive as a simulation tool and not a lighting/rendering tool. So I wrote various bits and bobs in PERL and MEL to munge the ribs into a series of bounding boxes (my script simply read the data and then created a maya ascii file with animated cubes to represent the agents).

Once the data was in maya I could then attach ribboxes to the various agents which would run the Massive DSO, sounds a little complicated but it just runs a program which spits out rib data at render time). This meant that no geometry was in the scene at any time. The shots were lit and rendered with renderman using custom prman shaders and hdri environments.

Midway through the project I was sent to LA to do the proper mocap shoot. This involved 10 days in total stuck in a warehouse working with and directing the mocap performers to get the results we wanted. It's fair to say that I learnt that planning is the single most important element that goes into getting good data from a mocap shoot. Plan exactly what you need and don't leave it up to chance. The guys we worked with out there (Giant Studios), Michelle and Tom, all the performers were really awesome and really made it work well.

We had dynamic LOD (level of detail) for the agents which meant that as the agent became bigger in camera space we could load in higher and higher resolution geometry in place of the lower res stuff. This worked really well and saved us a lot of rendering times. Renderman has a nice LOD feature which blends the different geometries together to avoid popping as the geo is switched out.

In the end we (the crowd team) completed about 30 shots of varying complexities: up to 3000 soldiers on screen, sometimes as big as 1/3 screen height (pretty damn close!), we also added loads of horses with riders. I was on the show for about 12 months most of which was spent designing the Massive agents and running the shots.

I don't want to bore you guys anymore so I'll leave it there ;) but if you have any specific questions or things you want to know about what we did please do ask. I have to add though I didn't work on the Lion or the Badger so don't ask me about them cause I don't know :D