One of the challenges I constantly faced while making ‘Vetri’ was that many of the scenes need plenty of characters. There is a lecture hall scene with many students, a wedding hall scene with many guests, a final exam scene with many students sitting in a classroom, and also at the end of the film, a stadium full of people. I’ll be tackling the stadium full of people scene with a different kind of approach, but in this article, I’ll detail how I went about generating a medium-level crowd for my animation film ‘Vetri’.
There were two issues I had to face when tackling these kinds of scenes. The first, motivation. Since the film is a one-man effort, generating and also animating all the different characters was going to be extremely time-consuming and slightly tedious. The other issue is the limitation of my PC processing capabilities. I’m using my trusty Dell Intel Quad Core PC with 8GB RAM and this can only handle at maximum 15 different characters before it crashes.To solve these issues, I had to cheat (a lot!).
For the first issue, I generated all the characters on-the-fly using the Cookie Flex Rig from BlenderCookie. This is the same rig that was used to generate all the main characters. I simply had to push a few sliders to give each background character variation. To animate each of the characters, I created a few basic loopable animation sequences (e.g, staring, looking around, bending up/down slightly etc) and applied them to each of the characters. I would also duplicate the same animation on a number of characters to save even more time and adjust the timing and scale in the dopesheet to look more random, and hence natural.
To address the second issue, I would first import only my main characters and animate them. I would then do the usual setting up of cameras and angles, finishing the animation and timing it, so it looked like the final animation, only without the secondary and background characters. In other words, at this stage, I would have a completed wedding scene, just no guests attending. Once I was happy with the animation, I would then look at each of the camera shots and angles and decide where the secondary characters won’t be seen. I would then save multiple copies of the same animation file and each file would render a specific shot from the scene.
I would then add the secondary and background characters per shot. This is shown in the figure below. Combining each of the shots together in the final film will make it look like the scene contains a very large number of people.
There were two issues I had to face when tackling these kinds of scenes. The first, motivation. Since the film is a one-man effort, generating and also animating all the different characters was going to be extremely time-consuming and slightly tedious. The other issue is the limitation of my PC processing capabilities. I’m using my trusty Dell Intel Quad Core PC with 8GB RAM and this can only handle at maximum 15 different characters before it crashes.To solve these issues, I had to cheat (a lot!).
For the first issue, I generated all the characters on-the-fly using the Cookie Flex Rig from BlenderCookie. This is the same rig that was used to generate all the main characters. I simply had to push a few sliders to give each background character variation. To animate each of the characters, I created a few basic loopable animation sequences (e.g, staring, looking around, bending up/down slightly etc) and applied them to each of the characters. I would also duplicate the same animation on a number of characters to save even more time and adjust the timing and scale in the dopesheet to look more random, and hence natural.
To address the second issue, I would first import only my main characters and animate them. I would then do the usual setting up of cameras and angles, finishing the animation and timing it, so it looked like the final animation, only without the secondary and background characters. In other words, at this stage, I would have a completed wedding scene, just no guests attending. Once I was happy with the animation, I would then look at each of the camera shots and angles and decide where the secondary characters won’t be seen. I would then save multiple copies of the same animation file and each file would render a specific shot from the scene.
I would then add the secondary and background characters per shot. This is shown in the figure below. Combining each of the shots together in the final film will make it look like the scene contains a very large number of people.
So far, I have managed to complete most of these medium-level crowd scenes without burning my CPU. For the final scene, I will need to create a large crowd to represent spectators at a sport event. I will need to make use of low-poly characters and instancing. I’m still learning how to do that and hopefully, when it’s time to animate that scene, I’ll be ready..