We recently created a scene that was populated entirely by CG characters. The scene is of a beach landing during WWII. Generating large CG crowds is known as Crowd Simulation, we wrote about this subject in our blog piece “Digital Character and Crowd Simulation”.
Crowd simulation and CG generally are brilliant tools in a filmmaker’s kit, but they have both strengths and limitations.
Previs is short for Pre Visualisation, this is a visual guide to the scene that the director can show to the actors, cast and effects artists ahead of time to give everyone a strong idea of the goal of a shot and have everyone operating off the same page.
Rigid objects are easier to create than organic ones. Humans in particular are the hardest things for computers to simulate. This has to do with the amount of detail we can read in human anatomy, emotion and expression, the way a person moves and rests will look jarringly unnatural if even slightly off.
However, even in these areas computers are now generating incredibly lifelike characters.
As we recently discussed in our blog “GLUE Looks at the use of VFX in film” for the film Terminator Genisys, MPC created a very successful fully CG render of Arnold Schwarzenegger.
The sort of attention to detail, cost and time associated with creating such realistic characters makes it impractical for use in large crowds. Crowd Simulation works brilliantly on groups of animals because we aren’t calibrated to understand their subtle body language and facial cues.
Nor in the majority of cases can we even distinguish between two animals of the same species.
In humans however, we read facial expressions and attach many complex characteristics to individuals, so when we see a crowd of human characters up close we expect to see them react emotionally to their environment through both their facial expressions and their body language.
Utilising Crowd Simulation
Our scene called for Crowd Simulation, but not photorealism. The objective of the exercise was to create a stylised and detailed scene.
Our character model is very simple and our scene has no textures, we realised that without a hugely detailed model and an unfeasible amount of hours of individual character animation we wouldn’t be creating a realistic scene.
Our scene therefore couldn’t be used as a stand-in for actors in film as it stands, however it would be a huge asset as a previs tool.
Alternatively, it could be used as an effect in the finished film, this would require some changes, most obviously in adding textures. Another change that could make a difference is for the CG characters to be supplemented with a performance by real actors, this would give the scene a grounding in reality that is key to holding the audience’s belief.
As can be seen in this breakdown from Hercules by Cinesite, Crowd Simulation is at it’s best on distant/background characters or when the focus of a shot is on a real performance from an actor that lends credibility to the effect.
Alternatively, this sort of scene is often used to great effect by Directors as a previs tool. A film that recently made use of this was The Avengers: Age of Ultron. If you’ve seen the final film you will know how much of this Previs Animation by The Third Floor made it into the finished version. This highly detailed animation conveys a much more accurate tone than could be achieved using storyboards alone.
The filmmakers can take a look at the previs animation and some concept images such as the one below and get a very good idea of how the final scene will feel to an audience. This can allow them to remove pieces that detract from the goal of a scene or the tone of a film, meaning that they can then go and shoot footage safe in the knowledge that time and money are not wasted on shots that they would later have to cut.