Category: Advanced and Experimental 3D Computer Animation Techniques

This blog post is about this next unit and our goals for our animation careers and expectations towards it, what we believe we need to improve and be better at.

Where do I start? There is so much I really want to be good at, I literally want to know how to do everything, but that is not a realistic view of the future. In this industry we are supposed to know a bit of everything but specialize in something orelse you’ll end up not good at anything, or at least not as good as you wanted to be.

So looking at the first term, I’m really happy with how I got better, but on my last 2 animations I noticed too many mistakes and I know why and how I did them. On the full body walkcycle animation I really struggled because of the spine but immediately after I asked Luke for feedback I noticed a real improvement mostly in the curves, now, on the body mechanics I wanted to do something more challenging and perhaps bigger than the others.

That was the animation work I was trully commited, I really wanted to have the model’s body do exactly as the reference footage and even changed somethings as a way to show myself I could also do things without reference – wrong! It wasn’t as good ofcourse and I made the mistake of animating every frame, not giving so much attention to the curves. I really liked the final result but having a keyframe on every frame makes it hard to make smooth movements, everything seems monotorized. Curve work is what I trully have to improve and get better at, as well as head and facial expressions.

On the last animation project from term 1, the performance animation, I took too much time looking for a good audio clip and then too much time recording myself, which was terrible and I ended up not really following my reference and just sort of performed while animating. Of course, it was a terrible decision and the lesson is learned! I’m never doing that again. I focused more on the lip sinc and the blocking had some mistakes, like the expression on his face was really dead and the feet were stuck to the chair.

Finally when making my showreel, I rendered most of my animations but I realised I don’t know much about it, as well as lights, all my renders looked a bit dead and that is something I want to change.

So what I really have to improve this term is:

• curves
• overlaping action
• facial expressions
• improve reference footage
• improve blocking
• rendering
• lighting
• camera work

In a few weeks we will be starting a new performance animation with two characters and one of the things I was keen to do is either someone fighting (verbally or physically) or some sort of sincronized dance…? It is still in the raw ideas department and everything needs to be considered, the next step is to find audio clips and choose 2 characters from the library.

The second video tutorial from Medhi was super fun as we had to create a particle simulation on a test model from Houdini that was already animated, make it look magical.

In the first video Medhi gave us an introduction to some concepts like: SOP – surface operators; OBJ – object context; DOP – dynamics context; ROP – render operator; VOP – vex operator. This week we started working with DOP and VOP as the DOP network was used for particles and VOP is the scripting language of houdini so it will be used for the simulation. A few more concepts were adressed this week related to the vex networks:

• i@test – integer -59687
• f@test – float -2.6587
• v@test – vector {1.15456, 2, -0.5}
• s@test – string “hello world”

These are the main values of language that vex networks work with and they are adressed with collors inside an attribute vop node, where anything can be added such as noise and color without having to actually know coding – as Medhi said. Which is great because I find codding very hard to understand.

So, this weeks task is about doing a bit of VFX and an introduction to particles, like before the first step was to create a geometry node, inside we had to add a node for the test geometry model, this test geometry is already animated so we can focus on the simulation and not so much on the animation.

For a better workflow the model cannot be time dependent as it is the input for the DOP network that is time dependent by default, so we have to create a time shift node and delete the expression it has on frames, so this way the animation doesn’t run while taking care of the simulation atributtes. After this, we create a POP network (that is actually a DOP network with a shortcut for a particle simulation network.

But before going in deep in to the simulation we need to look at how it is going to be read by geometry and for that there are 2 usefull nodes: pack and unpack. Pack node is used to read the geometry as one entity, so it will have no points, but unpack simulates all the geometry that allows us to see and manipulate it – so that is the one we are going to use for this task, before the time shift node. When these nodes are connected to the POP network, if we hit play particles start emerging from the surface of the model based on its vertices.

Inside the network, we can remove the merge node and it should look like this:

• 

• 

So, the pop object node works like a container for the particles, the pop source node are the particles and the pop solver is like a toolbox with everything needed to calculate the dynamics for particles.

Now what is needed is to add forces to the particles that are being created, so we can add a pop wind node or more than one and tweak them in order to have a more natural effect with multiple forces interacting with each other.

On the geometry node we should add a scatter node in order to control the amount of points we have on the models geometry from which the particles are going to be created from. Ofcourse this depends on how good our machines are because it is all cached to RAM memory, I cannot add that many points because my RAM memory is 16GB, but the result was equaly good in my opinion.

So after adding all the forces and connecting all the nodes this is what the DOP network looks like:

Inside the geometry node this is what is should look like:

A few more nodes were added such as the delete one that was used to delete points from which particles were emerging, this way the particles were only appearing in the geomretry that was facing up and the rest was deleted by normals. The copy to points node, along with the spere and atribute VOP nodes were used to make the particles, so the sphere was the shape of the particles and the attribute VOP was for their color, all this then had to be connected to a merge node so that we can have this amazing final result.

Term 2 is here and it began with a talk with Medhi a Houdini artist who is going to teach us Houdini for at least a month. I never worked with houdini but I’ve seen some things made with it that are just amazing, it is a 3D procedural software used a lot in the industry as it is so versatile and good.

I am excited to learn a new software and to really understand it, but the scary part is that everything is created with nodes, it can also be created on viewport but eventually everyone who uses houdini needs to mess with the nodes.

We met Medhi, and the day after we got a video tutorial explaining the software interface and a few experiences with some noise nodes to create rocks for example and with our first task – model a wood cabin! I always loved modeling, it was my first contact with 3D and it’s always good to know it.

I admit at the beggining I didn’t really understand much of the rocks exercise, nothing was really going how it was supposed to, so I just watched and took notes to begin doing the cabin. So we created a test model (for reference) and a geometry node and inside: a box and gave it the right size, then 3 more boxes that were the door and windows, have them on the right position and create a boolean node (just like maya) to have it shape the door and windows. After this we created a delete node that separated the roof and floor from the cabin structure to be possible to edit them separately.

It was a simple exercise but I learnt a lot, when creating different nodes we have a perception of the possibilities within the nodes, and in each one there are A LOT of options, which is amazing. The interface is super cool and I think… user friendly, for me at least! Some things are similiar to maya, some options and the way the software works, but takes a bit of pratice to know the place and the way some things work.

In the end this was the cabin:

The node network:

The task for this week was to add some more detail and some wood planks on the side of the walls and whatever we thought good for the cabin, using what we learnt in the session.

I finally booked my one to one with Dom!

It took a while to decide, as at first I thought about tracking the body of a man on a bike, but we did that right before the session so I wanted to do something different like a spaceship landing somewhere or some buildings emerging from the floor.

I went on pexels.com to search for footage and found a really cool one of a drone shot in the desert, a very open space with a ton of possibilities. With this footage I went looking for 3D models of buildings and spaceships, and found an awesome one.

I agreed with Dom to start tracking the shot before our session just to pratice a bit alone and to be quicker for the sessions, but I couldn’t export the footage with a decent quality to 3DEqualizer, every time I tried it the footage was full of noise!

So we began tracking on the session, starting for the floor on the left side, then right side “walls”, then floor and finally backgrounds, always tracking in parallel and perpendicular lines to be easier to have a perception of the points location. After this we select the floor points and create a plane on 3DE. We were able to track all the points in 1h40 and import it to maya.

• 

• 

• 

We were able to track every point in 1h30 and finished the project to import to MAYA. That really is the part I have more difficulties as what happens in my screen is always different. So, we imported the footage, and the goal was to make a polygon plane on the floor so we could import the futuristic car rig I found and animate it but I was not being able to do the same as Dom as in maya I was moving the footage instead of the plane I had created to place on the floor. It is a task for next session, next week!