Sunday, 7 October 2007

Ragdolls with Lip-Sync.

An idea I've been thinking over for having a skeletally animated mesh with the ability to lip-sync to speech.
  • Imagine a human ragdoll mesh. The head can be represented by a single body. If we remove the head (but nothing else, not even the neck), we still have a fully functioning ragdoll. It is important to note that the neck joint to which the head was attached should not be removed. Let us call the neck joint "NTAG".
  • Now imagine that the head mesh that we have removed is stored as a separate file in MD3 format. This MD3 file contains a single "tag", which is also called "NTAG".
  • When the ragdoll is loaded, the head mesh is loaded too and is rejoined with the body by linking the MS3D joint "NTAG" and the MD3 tag "NTAG".
  • Well, what have we achieved? Not much - we have the same model as before, only now the head is stored as a frame-by-frame model while the body is stored as a ragdoll.
  • Next step: For the MD3 head model, create a separate animation frame for every sound the human mouth can make. Each frame should be named appropriately (eg, the animation frame of the head making an "eeh" sound should be named "eeh", etc).
  • Using SDL_mixer, load some speech sounds into the project.
  • Hook up an SDL_mixer sound channel to an output preprocessor function. This allows the user to access any raw sound data for that channel just before it is played.
  • Analyse the speech WAV as it is played using this callback function. Whenever an "ahh" sound is played, set the MD3 head mesh to the frame named "ahh", etc.
  • To improve the effect, use linear interpolation for the MD3 frame transitioning to smooth the mouth movement and make it look more natural.

Progress on ragdolls.


I've made some good progress with my MilkShape-ODE Ragdoll project since I last posted here. Currently, the project can load a MilkShape3D model, find and load any associated textures and create a set of linked joints and bodies to form a skeleton. When the model is rendered, the positions of it's vertices are transformed based on the skeletal data.


ODE Bodies are simply points in space with a "mass" value. They can be joined together with other bodies using "ODE Joints". To illustrate how this works, imagine that your upper arm is one "body" and your lower arm is another, with your elbow being a "joint".

Bodies cannot collide with each other, as they are used to simulate dynamics, not collision. However, they can be paired with an "ODE geom". Geoms have no dynamics data (such as mass or inertia) but do have collision data. So if you were to create a bowling ball in ODE, the ball's body would be what gravity pulled down on - but the ball's geom would be what prevented gravity from pulling it through the floor.

Please note that ODE joints are not the same thing as MS3D joints. In ODE, vertices are associated with ODE bodies. Two ODE bodies can be linked together with an ODE joint. In MilkShape3D, there is no such thing as a body - every vertex in a model is linked to an MS3D joint. An MS3D joint may have a parent joint or may be independant. When an MS3D joint is moved or rotated, all vertices associated with that MS3D joint move with it (see my last post for more details on how MilkShape3D joints work). So, MS3D joints are roughly a combination of both ODE joints and ODE bodies.

How the skeletal data is generated.

The first step is to generate a list of ODE bodies. This is done fairly simply: the number of ODE bodies created for a model is equal to the number of MS3D joints specified in the mesh file. ODE bodies are positioned by adding the position vector of every associated vertex together and then dividing the result by the total number of vertices used, so they are generally positioned around the center of a "vertex cloud". Any ODE bodies which do not have any associated vertices are simply positioned at [0, 0, 0], to avoid a divide-by-zero error.

The next step is to join these bodies together using ODE joints. This is not quite as simple as generating ODE bodies. There are two reasons for this:
  • MS3D joints are joined to other MS3D joints, while ODE joints are connected to ODE bodies.
  • MS3D joints do not need to have a parent, whereas if an ODE joint is not attached to anything then it will join itself to the environment. This has the same effect as nailing something to a wall, because the environment is the world - meaning it doesn't move.
The answer here is to create an ODE joint for every MS3D joint that has a parent, join it to the ODE body with the same index value as that MS3D joint (so if you were working with MS3D joint #3, you would join the current ODE joint to ODE body #3) and that MS3D joint's parent MS3D joint's ODE body. This is not as complex as it sounds - it only becomes difficult because there are less ODE joints than there are MS3D joints.

Current work.

There are still a few things missing from the simulation yet. Firstly, the ODE bodies do not have any mass values set. I think that the best solution would be to assume a uniform density value for every body in the mesh (for example, 0.4) and then approximate the total area covered by the vertices of an ODE body using a cuboid or capped-cylinder shape.

Secondly, no ODE geoms are generated yet and so there is no collision. Originally, I was planning on approximating ODE collision geoms in the same way as I would approximate mass but, while I can get away with comparatively slight inaccuracies in mass approximation, collision errors are more "visible" to the end user. My current thoughts are to generate a separate trimesh for each ODE body in the mesh. This would be very accurate, but it could be computationally expensive.

Possible collision problems.

Neither of the above approaches really solve the problem of triangles which span vertices which are associated with different bodies. While most triangles in a mesh belong to a single ODE body, triangles which are used to join the vertices of two ODE bodies together (such as "elbow triangles" which stretch from the upper arm body to the lower arm body) will not have any collision data.

Possible solutions are to create an ODE sphere geom for each ODE joint or to generate a trimesh for the entire model every time it changes. The first approach could lead to redundant ODE geoms being created for joints in locations such as at the top of a string, several feet away from a "puppet" model below and would also require me to find a way to guess what the radius of the sphere should be. The second approach could have issues if it interferes with the "temporal coherence" data that ODE uses for trimesh collision.

Sunday, 16 September 2007

Skeletal Animation and Ragdoll Physics using ODE and MilkShape3D


I haven't got around to working with skeletal animation or ragdolls before, though I am aware of how the theory works and I am familiar with how it could be implemented using the Open Dynamics Engine.

In addition, when working with 3D models, I have always used Quake 1 .MDL, Quake 2 .MD2 and Quake 3 Arena .MD3 files. These formats use "mesh deformation" or "frame-by-frame" animation techniques which make them unsuitable for ragdoll work.

For those of you who are not familiar with the terms "mesh deformation" or "skeletal animation" I offer this explanation:

Mesh Deformation

In "Mesh Deformation", a model is animated in the same way as a flip-book. A flip-book animation of a running stick figure would consist of, for example, 5 different drawings of the stick figure, each one with it's arms and legs in slightly different positions, so that when the drawings are cycled through quickly it appears that the figure is moving. This is why "mesh deformation" is also known as "frame by frame" animation - a 3D model which uses this technique would actually consist of several slightly different models (in the same way that the stick-figure consists of several slightly different drawings) but only one of these models would be shown at a time (in the same way that only one stick-figure drawing would be visible at any precise moment). The model would cycle through these separate frames of animation much like a 3D film, creating the illusion of movement.

This basic technique has a slight problem, though - if the frames of animation change too slowly then the model will appear to be moving jerkily, as the user is able to make out each frame individually, rather than see the overall "effect" of the frames as they change. A perfect example of this is in the original Quake 1 game, where monsters move in a rather jerky fashion no matter how high-end your computer is in terms of graphics power.

There are two solutions to this:

The most obvious approach is to simply use more frames of animation, more rapidly. This means that, for example, the running stick-figure now has 10 frames of animation rather than 5. This allows the artist to create "in between" animation frames for the animation so that the movement of their stick-figure looks more fluid. This technique can be used for both flip-book animations and 3D meshes. However, this approach not only requires the artist to do a lot more work but a (drawing/model) with 10 frames of animation will require twice as much (paper/computer memory and disc-drive space) than one with only 5.

The alternative, used in Quakes 2 and 3, is to use "linear interpolation" (also known as "lerping"). This essentially means that, if a game engine should show animation frame #1 at time=30 seconds and animation frame #2 at time=34 seconds and the current time is 32 seconds, the game will work out a new position for every vertex in the model, exactly halfway between the positions specified by animation frame #1 and animation frame #2. The artist is only required to create a few "key frames" for the model's animation and the grunt work of creating "in-between" animation frames is handled automatically by the game engine. Even better, as this work is done "on the fly", the model will not only transition smoothly between frames 1 and 2, but frames 4 and 9, 7 and 3 or anything else - useful in games, where you may want to go from a running animation to a death animation, a running animation to a shooting animation or a running animtion to a crouching animation. This is the same technique used by Adobe Flash for animating shapes, although Flash refers to it as "tweening" (as it is used to do the inbe-tweening of key-frames).

Skeletal animation.

"Skeletal" (also "jointed" or "boned") animation is a truly distinct concept from that of mesh deformation.

First, a series of "joints" is created. A joint consists of a unique joint ID (such as a number or a name string), a parent ID (similar to the above, but instead of referring to itself it refers to the unique ID of another joint), an XYZ position in 3D space and an angle (probably stored as a quaternion or a matrix).

If a joint does not have a parent - for example, joints may be numbered starting from 1 and the parent ID of the joint is set to zero - then it is completely independant of external influences and it can be positioned or rotated however it wants to be. If the joint does have a parent joint then it's position and angle will be influenced by the position and angle of that parent; It can be referred to as a "child" joint. An imaginary straight line, connecting a parent joint to a child joint, is called a "bone" and a collection of bones forms a "skeleton".

To illustrate, imagine that you have one "parent joint" positioned at your shoulder and another "child joint" positioned at your elbow. When your shoulder joint is rotated, your elbow joint is moved - but when you rotate your elbow joint, your shoulder joint does not change position. Now imagine that your enitre body is "jointed" in this way. Every joint in your body can be represented as either a parent joint, influencing others, or a child joint, being influenced. Some joints may even be both - your elbow may be the child of your shoulder but it is also the parent of your wrist. Additionally, parent joints may have multiple children - for example, one wrist influences five fingers.

Now that we understand how joints can influence each other, we can go on to explain how models can use jointed techniques in animation. Once an artist has created a skeleton for a character, they can create polygons around the bones as "flesh". Every vertex in the model is then mapped to a joint, so that, when the joint moves or rotates, the vertices will move along with it (in much the same way that a child joint will move along with a parent joint).

The result is a model with a skeleton that acts in a similar manner to that of a human. The model can be loaded into a computer game and the physics engine will be able to simulate forces acting upon the model's skeleton in the same way that forces such as gravity influence our own skeletons in reality. Combine this with collision detection and it becomes a "ragdoll" of the kind seen many recent video games.

More recently, the idea of using a system of "weights" to provide more realistic skeletal animation has become popular. However, this is not an area into which I have done much research as of yet.

"Tagged" Animation

To be completely fair, Quake 3 .MD3 meshes do allow a basic form of "jointed" animation (in addition to the legacy frame-by-frame technique which they have inherited from the earlier MDL/MD2 formats).

Each player model in Quake 3 consists of 4 parts: A "head" model, an "upper body" model, a "legs" models and a "weapon" model. These are stored as separate .MD3 files, each of them containing a list of "tags". A "tag" is simply an XYZ position, an angle stored as a 3x3 matrix and a name string (which is used as a unique identifier). This makes them almost, but not quite, the same thing as joints, as they do not have "parent tags" - each tag is influenced by nothing but itself. To connect the head model to the upper body, both the head and the upper body files will contain a "neck" tag. When the models are loaded, the head mesh is repositioned so that it's "neck" tag has the same position and orientation as that of the torso "neck" tag.

There are disadvantages of this approach:
  • As I have already mentioned, tags cannot have parents and so skeletons and bones cannot be formed.
  • Models must be scattered across several files rather than grouped into a single, self-contained one.
  • Triangles cannot use any vertices which are stored in seperate files. This means that each mesh file will appear to be physically unconnected with it's fellows.
It is worth noting that id Software (the creators of the Quake series) were originally going to use skeletal animation for .MD3 meshes, storing animation sequences in separate .MD4 files. However, this was never finished and both the .MD4 file format and Quake 3 skeletal code remain incomplete (this is why Quake 4 meshes are called "MD5" - to prevent confusion with the half-finished MD4 skeleton specification).

The Project.

My current project aims to allow any skeletal mesh formation to be loaded into my physics engine and be treated as a ragdoll. Unfortunately, models stored in Quake .MDL, Quake 2 .MD2 and Quake 3 .MD3 format do not contain any skeleton data and are thus unsuitable for ragdoll models, meaning that I cannot re-use my old mesh loading code. Instead, I have opted to use the MilkShape3D binary ".ms3d" format. It is a very simple format and yet it contains all the information needed to generate a ragdoll and render it.

At time of writing, I have finished creating a simple .ms3d loading API written in C (which loads almost everything aside from the joint animation data, which is not needed as I only require the joint's positions and parent data) and I am writing a simple test application using ODE, OpenGL and SDL.

As I am not a skilled modeller, I have decided to borrow the "Angelyss" model from OpenArena and create a skeleton for it rather than create a whole new character from scratch. So far, I have imported the .MD3 file and created a skeleton but have not yet finished assigning vertices to joints. Click here to see the image.

Thursday, 13 September 2007


For a while now, I've been meaning to create a website that I can use both as an on-line portfolio and as a “learning diary” for my experiences in the world of programming. I originally tried to do this with a variety of pure HTML sites but these were cumbersome to update and maintain - although I learned a lot about HTML by creating them, so I suppose that it was a worthwhile endeavour overall (it is worth noting that, due to a no-PHP restriction with my freely-provided University web-space, my site there is still written entirely in pure HTML).

My next thought was to create a PHP-based site, using an SQL database to store articles. I therefore installed Apache, PHP and MySQL and set about learning how to use them all. I think that I learned a great deal on this brief project but I eventually decided that a Blogger account would be a better idea. This was mainly because creating a Blogger account costs nothing, whereas a decent amount of web-space and bandwidth with PHP and MySQL support can be costly, especially if you want to buy a domain name as well. The Blogger system would also allow features such as user comments, which I had not originally thought about but which would undoubtedly be useful in getting feedback and advice from visitors with more experience than I in whatever field I decided to research next.

What can you expect to see on this site? The majority of my postings here will be programming related, relating either to basic coding topics such as successful encapsulation with object-orientated techniques and programming correctness or to higher-level, theoretical concepts such as flexible-yet-secure game engine design and AI techniques. For anyone who is not so very interested in the raw work that makes games and applications operate, I shall be posting images of rendering experiments and possibly even the occasional downloadable demonstration program every now and again.

For anyone wondering where the title of this Blog comes from (no, it is not my real name), you might like to know that I've always been interested in the work of id Software – including two of it's founding members, both of whom have now left.