7m836 Animation and Rendering
Thomas Krijnen 0590144
Assignment 3: The movie
Starring
Jodie
Jodie is the star of this short movie and has already featured in many productions before. After all, she’s freely available at http://gfx-3d-model.blogspot. com/2009/12/girl-3d-model.html
Guest starring
Teapot
No CG animation can exist without a teapot. For that reason a very special teapot fulfils a guest role in this movie.
Introduction
The movie takes place in a building that has been designed in the simultaneous course of Freeform Design. The script of the movie is a short story alongside a walkthrough of the interior. The building itself is modelled in Rhinoceros. All the other modelling steps have been performed in 3ds Max. Some stills are featured in this report, the entire movie has been included on the accompanying CD. Preexisting renderings of the building interior have not been made. Therefore it is impossible to take existing renderings as a starting point for an evaluation of the spatial qualities in relation to rendering techniques. The images are rendered with the 3ds Max Scanline renderer. In the Blender Sequencer the frames are combined into one video file. Some minor editing has been applied in the latter to the ending of the film. It is impossible to describe all aspects and techniques used in this film, but in the following pages several crucial techniques in relation to the contents of this course will be highlighted.
Screenplay
7m836 MOVIE
INTERIOR SCENE - Inside 7m842 Exposition Building
The viewer moves in the longitudinal direction of the building when, through one of the openings, the viewer suddenly gets a glimpse of Jodie walking in the distance. The viewer starts following here, just able to see her turn around the corner. She moves into a location that seems to be a dead end, but somehow disappears. The viewer looks around seeking for clues when in the corner of it’s eye suddenly a teapot on a table has appeared. It evens seems to move when one looks away. Then out of nothing the teapot walks away. The viewer follows the teapot, again not able to really catch up. When also the teapot has disappeared, the viewer notices the beautiful exterior of the building and starts moving towards the window. The movie fades to white and ends.
Objectives
The most important aim of the movie is to create a mood that is specific to that of a dream. Most notably, reflect on situations in which the dreamer is aware of the fact that he is dreaming and knows certain situations are not real, but still is unable to get in control of what one is dreaming and is carried away by its own subconsciousness. The inevitability with which things happen without the possibility for the dreamer to interfere is a situation similar to those in which someone watches a movie.
The movie also reflects on meaning. Does this movie have a meaning? Do dreams have meanings? Do actions of other people in real life have meanings? For the viewer of this movie it may seem as if the teapot wanted the dreamer to follow him and eventually see the exterior. This way meaning is attributed to the behaviour of a teapot, a ridiculous assumption in itself. The same holds true for the girl, one assumes she was aware that the viewer followed her, but then again, never directly looked at the viewer.
Several aspects of the movie reflect on dreaming. The white interior has a surreal feel to it, it is slightly enlarged giving the scene an additional unreal quality. The uniform lighting levels, enhanced by the radiosity calculations and the diffuse white materials, enhance the surreal felling. The lack of detail in the interior also resembles that of a dream, since in dreams, one would only notice details with an impact on the development of the dream. The floating movement of the camera, with no real relation to the amount of time it would take in real life to cross such distances, also contributes to this feeling. Of course, so do the characters, the girl maybe literally the girl of our dreams, and walking teapots are not that common in real life. To simulate a hazy kind of environment a Blur effect has been added to the rendering pipeline. In the ending the viewer seems to realize he is trapped inside a building with no entrance or exit whatsoever and then wakes up.
The next pages show stills from the movie of the key scenes.
Assignment 3: The movie:
The making of
The downloaded model
Animating Jodie
The model, as it was downloaded from the Internet, was not animatable because it had no skeletal structure, i.e. it had not been rigged. Furthermore it had a quite articulate pose that needed to be addressed.
For the skeletal structure the choice has been made to use the biped system. The biped system is a hierarchical system of bones that, in its default state, mimics the anatomy of the human body. By linking the downloaded model to the biped system, the polygonal skin can be animated by transforming the bones.
The model in a neutral pose
The influence that the bones of the biped have on the polygonal skin is based on the distance between the bone and the vertices of the skin. Therefore in order to prevent vertices of the torso to move along with the arms, the arms had to be moved away from the torso and pelvis. Also the legs had to be moved to a more symmetrical state. This has been accomplished by using translations and rotations of individual vertices with Soft Selection.
With the biped and the skin in roughly the same pose, the Physique modifier is applied to the polygonal mesh. The modifier tracks the transformations of the bones
The 3ds Max Biped in the same pose
and deforms the mesh accordingly. Several adjustments had to made to the default settings in order to make the mesh transform correctly. Because the model wears high heals the influence region of the foot bone had to be increased, as well as the influence of some of the other bones because of some inaccurate positioning of the skin in relation to the skeletal structure.
The animation itself has been accomplished by using a technique called footstep animation, by which it is possible to derive all transformations, in the space and time domain, of the entire body, solely on the placement of footsteps.
Animating with footsteps
In order to derive the position and orientation of all the limbs in the body from the footsteps, an advanced system is used that can be explained in terms of inverse kinematics and forward kinematics. Inverse kinematics is used to derive the position and orientation of the joints in the legs from the position of the footsteps. From the pelvis upwards, forward kinematics is used to mimic the balancing motion humans make when walking. These include the gentle swinging of the arms, the twisting of the torso and a slight nodding of the head.
Walking teapot model
Animating the teapot
In contrast to the biped system, for a walking table with a teapot 3ds Max had no ready made solution available.
The bone structure itself is very straightforward. It consists of three bones for each table leg, with the last being the foot bone. The foot bone does not play a role in animating this particular model, but is placed anyway in order to be able to use the IK Limb Solver. This IK solver is suitable for solving for the legs and arms of human characters (and in this case also of tables) and requires a setup with three bones.
Bone structure
With the IK Limb Solver, the degrees of freedom of the joints are automatically correct for the shoulder, elbow and wrist joints. A setup that also suits the joints of this table model. Another advantage of this IK Solver, is that its solution resemble those of actual human movement, it tends to bend the knee earlier than strictly necessary, and in this way resemble the walk cycle of humans.
Animating this four legged character proved to be quite difficult, partly because the lack of real life reference material. Setting up a good hierarchy was beneficial for gaining plausible results.
Walking teapot hierarchy
For ease of selection and manipulating keyframes in the Dope Sheet, the entire system is grouped into a single object. This is the first level in the hierarchy. The teapot and the bones are parented to a Point Helper called Center Point. This has been done to aid moving the character as a whole. The mesh of the table itself did not need to be part of this hierarchy, since its deformation and movement is derived from the bones by applying a Skin modifier. The four goals of the IK Solvers are also not parented to the Point Helper. This enables the table to bend its legs simply by moving the Point Helper downwards.
The pivot point of the top bones would then be moved downwards and the IK solver goals would remain on the ground. The steps are animated by keyframing the position of the IK Solver goals. A single step has been animated for each table leg, these are duplicated using the paste relative command in the Dope Sheet to obtain the full walk cycle.
Rendering of the interior
Lighting strategy
The lighting of the geometry of the building comes from a single hemispherical skylight that is propagated through the interior using a radiosity calculation. The uniform lighting of both the skylight and the radiosity calculation contribute to the serene feel of the movie.
Since the animation consists of over a thousand frames, the view independent radiosity solution has been made of very high quality, reducing the need for view dependant regathering. However, this solution has lead to distinct artefacts where the walls and the
floor intersect. Two solutions exist for this problem. One is two use the regathering of indirect light when rendering the radiosity, this was no option because of the dramatic impact on rendering time. The second would have been, to make sure the vertices of the walls connect to the vertices of the ground plane. It would have been possible to make such a ground plane in Rhinoceros. But in fact, these distinct darker areas proved to be an improvement of the visual quality of the renders. It outlined the intersection between walls and floor and gave more definition to the freeform surfaces. In reality such an outline could also exist
Soft shadows independent of the radiosity solution
because these are areas where dust accumulates or because of the finishing process of the floors.
The movement of the animated characters would normally invalidate the radiosity solution since geometry would have been moved. Therefore these moving characters are excluded from the radiosity calculation. They are lit by an independent light source, a directional light. This light also casts a shadow on the ground floor. 3ds Max does not offer a Shadow Only light, as most other packages offer, so in 3ds Max it is not possible to cast shadows on a surface without illuminating it.
In order to get shadows without additionally illuminating the floor surface on top of the lighting levels from the radiosity solution, the additional lighting of the directional light has to be compensated by an additional identical directional light with shadows disabled and a negative intensity. Only lighting the moving characters by this single directional light is not sufficient, but introducing ambient lighting is also not an option because it conflicts with the radiosity calculation. To obtain an effect that is similar to ambient lighting, the self illumination slot of the material definition is used to match the lighting of the moving characters to the radiosity of the interior.
Looking up by moving the camera target
Camera strategy
A technique that is commonly used in architectural visualisation is a Shift Lens. Shift Lenses exist in real life but are extremely expensive, therefore they are used more often in CG.
Especially for the visualisation of interiors it is aesthetically pleasing for vertical lines to remain parallel. To emphasize the high ceiling of this interior space the camera constantly looks up a little bit. This rotation around the local X axis of the camera and the resulting vertical perspective make vertical lines converge to a single
point. This makes the space more dynamic and more difficult to read, it does not fit the serene atmosphere this animation tries to spread.
3ds Max does not offer the possibility to use a shift lens directly, but gives two ways to indirectly get the same effect. The first is to apply a Camera Correction modifier to a camera with an elevated target. This method is somewhat unpredictable, because of the reversed method and the automatic guessing of what to correct. A more intuitive way is to apply a Skew modifier to a camera with no elevated target.
Looking up by applying a Skew modifier to the camera
The skew modifier elevates the far plane of the view frustum, but keeping it parallel to the near plane. Thus maintaining parallel vertical lines while looking up.
Unfortunately, the use of this technique did not turn out fully integrated in the 3ds Max rendering pipeline. The grass in the exterior view has been made using the Hair and Fur modifier. It is great way to render animated photorealistic grass within reasonable render times and memory use. It is rendered in a separate pass after rendering the other geometry. Unfortunately it does not take the shift lens techniques described above into account and renders the grass as if
the Skew modifier was not in place and the camera target was elevated in the usual way. To work around this undocumented bug, a transition has been made at the time the walking teapot is introduced. The camera skew amount is gradually decreased and after that the camera target is gradually elevated. When comparing the exterior view in the images on this spread, one sees the superior compositional effect and parallel vertical lines one yields with the shift lens, and one sees the displaced grass.
The camera is animated by simply keyframing positions of the camera and target over time.