The basic premise is to build a short 2 minute animation involving a hamster in a ball and a soccer game. A little crude but I hope that it will end up being a fun video.
- A hamster is sleeping in a clear ball
- Pan out to see that the ball is in the middle of a field
- Ball is kicked, waking the hamster and causing panic within the ball
- The hamster runs haphazardly while pursued by players
- Hamster escapes by scoring a goal
These are the technical details that I will need to overcome to complete my project:
- Building the human model for the players
- Building the hamster model (this will be require the majority of the work)
- Animating the hamster in the ball
- Animating realistic human movement
I wil be using the Blender 3d Graphics program to build my animation.
Even though the assignment is due in week 12. I've given myself an additional 2 weeks for poor estimation/other craziness.
- Locate human model (from my research there are a number of open source models of humans available)
- Begin building hamster model
- Complete hamster model
- Begin basic animations of hamster (running and sleeping)
- Begin human animation
- Continue work on animations
- Complete animations
- Build scene
- Combine animations
- Complete animation
I spent most of week 4, researching animation and modelling. I went through this tutorial that will prove useful when animating the hamsters running across the ground.
Found a bear rig, that I will try to convert into a hamster. There was also a mouse rig but it looked less like a hamster. However there is currently no rig set up for the face so I will have to mock something up for that.
Most of this week was taken up doing the flame assignment. I did find a good model for a human that should suit my needs for the soccer players. The players will all look the same but I don't forsee many close up shots of the players. So this shouldn't be a problem.
Animated the start where the players shake hands and the camera zooms into the ball. Animated the ball being kicked into the air.
Animated the player kicking the ball. Since the model I am using is quite simplistic, it is quite difficult to get the right and angle and flow of the player. Although the initial kick looks reasonable, the follow through doesn't look quite right. The trajectory of the ball also needs some fine tuning.
Found a model of a squirrel that I can use as a replacement of the hamster. I have animated the squirrel waking up and the squirrel running. This website has proven invaluable in developing a realistic run cycle. Although it is not blender specific, it has pointed me to the features and movements I need to concentrate on and where to start.
Built the opening sequence and made further modifications to the running sequence to make it look more realistic. I originally had the opening sequence zoom in from above, however its was hard to get a good idea what was going on from that angle. So I have changed the camera to zoom in from the side.
Built the squirrel animation as it is kicked along the ground. Modelling the bounce of the ball has proven to be quite difficult. For rendering the ball, I have used the values from this website as a guide.
Found a useful tutorial on how to model grass, using blender's in-built particle system. http://blenderartists.org/forum/showthread.php?t=68269. Blender allows you to specify the number of particles used to render the grass. A good number appears to b 100000, although at that level it slows down my computer considerably.
I started rendering the animations, for the more complex animations one frame has taken 20 min to render. Without a render farm at my disposal I won't be able to render some scenes with ray tracing and high quality anti-aliasing.
I found a really good tutorial on how to create a realistic looking LED billboard. I also used this http://cogfilms.com/tutorials/Animated-Procedural-Textures.pdf%20tutorial in order to get the texture moving realistically across the sign.
The different scenes that I ended up building were:
- Starting sequence (zoom into the ball)
- Squirrel waking up
- Ball kicked
- Ball bouncing and then stopped
- Squirrel stunned and gets up
- Squirrel running through a sea of legs
- Squirrel "scores" goal
The individual scenes were put together using Windows Movie Maker. This turned out to be more painful than anticipated (see final thoughts).
In order to get a full animation rendered I have had to compromise on some key features.
- Some scenes won't have a ball, the ball being transparent eats up the majority of the computation time, so for particularly complex scenes (like running) there will be no ball.
- Some scenes won't have raytracing or shadow maps enabled. This makes the scene look a little less colourful than wanted.
- No background, I had originally planned to have trees and spectators in the background but I never got around to it.
- I have had to implement limited motion-blur for running scenes. Since motion blur requires a single frame to be rendered multiple times, I couldn't dial up the motion factor as much as I would have liked. Thus the squirrel appears to run quite slowly.
- Clothes: In the majority of the scene, the players are naked. I just didn't get around to putting clothes on them.
- Fur and Tail: The original model has textures that you can apply to the squirrel to give it fur and a bushy tail. I never got around to adding these textures.
After completing the animation there a few things that I have learnt from the experience:
Render Farms: One aspect that I severely underestimated was the length of time needed to render a frame. Although in hindsight it is fairly obvious that rendering each frame would take several minutes with motion blur and raytracing, this was something I had not considered. For a full-blown animation a render farm would be essential
Transitions: When I had finally put together all the scenes I realised that I hadn't really planned out how each scene would transition into the next. Carefully planning this out would have made my animation flow much, much better.
Speed: I have noticed that it is very hard to judge the speed of the animation in Blender. So that an animator can fine tune each movement, Blender slows down the animation in playback mode. Thus when the scene is fully rendered it can be too fast. I tried to compensate for this, but then ended up with an animation that was too slow.
Post-production: In order to fit all the scenes together I used Windows Movie Maker, as it was already available on my machine. Although this program is great for compiling home videos it just doesn't have the features needed for post production (such as slowing down or speeding up a scene). It also could not really handle the size of files that I was giving it and would crash regularly.In hindsight, I should have researched video editing software as well.
Blender: Before starting this project I hadn't used Blender before so this was a very useful experience. My thoughts on the software are that it is extremely powerful and has a thriving open source community that makes it relatively easy to pick up the basic concepts. However I have found the animation of models to be extremely complex and time-consuming although not necessarily hard. I think the most use can extracted by developing python scripts to do the hard labour for the user, this was something I didn't have a chance to investigate. There are also far too many parameters in Blender for one to get any real understanding of the software in this space of time, but the plethora of tutorials help immensely.