I have been working on and off on the cheetah model for my major project since before christmas, so I thought it was about time I posted it here 
There is however something with it that I’m not quite happy about, especially with the head, but I can’t put my finger on it. Well I have to finish it off soon, so there will hopefully be an update within the next week.
For my Innovations Project I did a painterly motion blur for moving images.
I got inspired to do this from images by the artist Bill Hall.
I started by looking into ways of manipulating motion vectors, rendered from a short clip of animation. I tested different masks made from both moving and still images, such as smoke and elephant skin. The idea was to mask out areas that wouldn’t be affected by the motion blur and therefore get an interesting pattern for where the detail would appear. I also tested different inputs to change the shape of the motion vectors. On its own this turned out to give a too random result.
Later tests were done with displacements. I started out by doing tests on still images by experimenting with distortion filters in Photoshop. By following an article in Computer Arts Magazine by Pete Harrison, Photoshop Displacement Effects, I produced the following images.
Original image displaced with a black and white image. (Black and white image from Pete Harrison.)
Original image layered with the displaced image where parts have been erased.
I later tried to get the same results for moving images by using moving images as input for the displacement, in Shake instead of Photoshop. I tested a number of different pieces of footage as displacement, both 3D and video. The conclusion was that high contrasted footage, such as coloured liquid on dark backgrounds, gave more of a trailing effect than smoother images, such as smoke. I also noticed how much easier it is to create something that looks nice for a still image rather than moving ones.
A test where two pieces of footage of different kinds of liquid (filmed) were used as displacement in two layers. (The middle image is from footage shot by students on the MA Digital Effects Course.)
Mixing and layering different tests lead to the latest image sequence.
In the following 5 steps, Maya and Shake have been used, though other 3D and post production software with similar functions can be used.
Step 1 – Animated sequence
To create this kind of motion blur, an animated sequence with a subject moving relative to the camera is needed. The images generated from this sequence should not have any kind of blur due to motion (other kinds of blur, such as depth of field, are acceptable). This is the base pass.
Step 2 – Exaggerated “natural?? motion blur
This step can be done in one of two ways. The second pass can be rendered out with motion vectors (using Reel Smart Motion Blur). Motion vectors allow you to interactively change the amount of motion blur in post production software (assuming you have the Reel Smart Motion Blur plugin). This makes it easier to see how much “natural?? motion blur is needed to fit with the later passes, without having to render out more than the base pass (image below).Exaggerating it slightly will give a more dramatic effect and will enable this pass to fit better with the subsequent stages of this process. Alternatively the animated sequence can be rendered out with “natural?? motion blur already added. However, bear in mind that if this pass does not fit with later passes it will have to be rendered out again with different motion blur settings.
Base pass combined with motion vector pass giving, with the right settings, the exaggerated “natural?? motion blur pass.
Step 3 – Displacement with irregular background
The next pass is generated using an IDisplace node. The IDisplace node has two inputs. The first is the image to be displaced and the second is the source of the displacement. The IDisplace node distorts the images of the first input in x and y according to the colour channels specified of the source of the displacement, the second input. Using different colour channels for the distortion in x and in y can create a slightly asymmetric distortion.
The first input used for this pass is the base pass. The second is a filmed grainy irregular background, where the red channel is specified to distort in x and the green channel in y (image below). This creates a slightly irregular distortion of the base pass that works well as a merger of the passes produced in previous and following step.
Base pass displaced with a grainy image as source.
Step 4 – Displacement with filmed liquid
This step also uses the IDisplace node. First create three instances of the base pass fileIn node, A, B and C. Leave A unchanged, but reposition nodes B and C by -1 and -2 frames respectively. Nodes B and C now appear behind A as a trail when the sequence is played back. The fileIn nodes are then plugged in to the first input of three separate IDisplace nodes.
The source of displacement is footage of a liquid running down a pane of glass. Depending on the quality of the footage, this fileIn node may need to have its contrast increased.
The filmed footage should be rotated and skewed to fit the direction and the perspective of the object in motion (image below). The output is then plugged into all three secondary inputs of the three IDisplace nodes and again the red channel is set to distort in x and the green in y.
Base pass displaced with filmed footage of liquid. Here the three displaced images are layered, with different opacities, into one.
Step 5 – Mixing and layering the different passes together
Now the different passes can be added together. First mix the exaggerated “natural?? motion blur pass of step 2, with the displacement pass of step 3 and leave the opacity at 50% to show the two passes equally. To achieve a blurry effect on top of this, layer the output of the mix layer with the “natural?? motion blur again in a max layer.
Combine B and C with a second mix layer, B as input one and C as input two. Use this output in a third mix layer together with A. Adjust the opacity of these two mix layers so that A is more visible than B and B more than C, to give an effect of the displacement fading away when played back.
To complete the last step of creating the painterly motion blur, combine the output of the max layer and the output of the third mix layer in a new max layer.
After finishing the woman’s rig I started rigging the second character, the cheetah. Since I had never rigged a four legged animal before, I had a look around earlier in this project for examples of how other people have rigged quadrupeds. I found one free rig of a puma by Kiel Figgins (Rig Set 5), that I used in my block test. I first thought I could use this one even in my final piece, but fitting it to my character. However, from using the rig in my block test I had found a couple of things that I wanted to change. When I started changing these, though changing these made me find more things that I wanted to have in a different way. I therefore ended up making my own rig, but following Kiel Figgins’ puma as an example.
The Skeleton – In opposite to the human rig, where the root is close to our pelvis, with spine joints working up towards the head, the root joint of the cheetah is between its shoulder blades, with four spine joints going down towards the tail. The fourth of these spine joints is treated as a second root, to allow for the front and the back of the cheetah to move independently. From the back root the back legs follow with a similar setup to a human leg, with hip, knee, heel/ankle, ball and toe. The difference is that a cat’s ball is much more flexible than the ball of a human foot, which allows the cat to walk comfortably in away that would be equivalent to us walking around on our toes. The heel/ankle of the back leg is therefore placed much higher than on a human leg.
The tail follows after the back root, similar to an elongated spine.
In the same way the back legs has a lot in common with human legs, the front legs are similar to human arms, with shoulder blade/clavicle, shoulder, elbow and wrist. After the wrist cats have paws instead of hands.
The Back Legs – To get the movement of a cat’s leg I have used three IK chains on each back leg. One from the hip to the heel, with much the same function as the hip to ankle IK would have on a human rig. From the heel there is a single chain IK to the ball and one from the ball to the toe. Parenting the first IK handle to the ankle control and the second and the third to the paw control makes you able to split the movement so the heel and paw can be moved independently. Since the heel of the cheetah works in a similar way to a human’s heel, it should only be able to rotate in one axis. Therefore to not “break” the heel when the ankle control is moved from side to side, the movement of the paw control is driven by an expression that always keeps the paw in line with the ankle.
The Front Legs – The clavicle of the cheetah is placed close to the front root, to get motion of a shoulder blade. These are then controlled by FK Shoulder Controls.
Similar to the back leg setup, the front legs are also set up with three IK chains parented under two controls with the same expressions.
The Body – The spine is split up in two independent roots by using two IK chains. The first is a spline IK starting from the front root and ending on the third joint down the spine. From this joint another IK starts, a single chain IK, ending on the back root. This gives you a lot of control over the curvature of the back and you can also move the back part of the body independently from the front.
The Tail – This part of the cheetah is controlled by a spline IK with control clusters assigned to the control vertices of the curve of the IK. I tried at first to just use FK controls for the tail, but decided against it since it requires a lot more controls for the same round movement as with a spline IK.
The Head and Neck – I wanted to be able to have isolated head movements, the head and neck are therefore setup in much the same way I set up the head and neck of the woman, following Jason Schleifer’s tutorial. To get complete control over the ears they were set up with FK control curves. The eyes were also setup in a similar way to the woman’s, with aim constraints to get isolated motions.
Gave the woman some clothes and hair a while ago, but never posted them. So here is a small update on the female model.
I plan to use wire deformers for the dress with hair dynamics and I probably have to look over the positioning of the eye brows again. I know the hair line isn’t too pretty atm either, but I will see how much I can get away with in terms of texturing before I change anything.
A short one today guys (in both footage and description):
Right-click and “Save As…” to download dl_mp_animDaily_004.mov (1.62mb)
Description: Tweaked entering the apartment and added to the crime scene arrival.
I had a fairly slow day yesterday which now means I’m behind schedule slightly but I should have this done in time for my tutorial tomorrow. Oh yeah and the lights are a little broken - I slowed down his animation a bit and changed the camera timings, so now the “cheating lights” come on a little bit late - I shouldn’t need to point it out, you’ll definitely not miss it ;).
For this clip, I was trying to push the contrast even further and generally darken it down quite a lot. I was just looking at the frames I rendered overnight and got worried that I’d pushed it a little too far. The .tiff frames are quite difficult to make out but for some reason they’re much lighter after they come out of Premiere. Not ideal but at least I don’t have to re-render for today’s update. Check out how big the difference is:
Anyone know why this is?
I think I’ll tone the lighting down a tiny bit, but I have to try and stop it becoming more important than getting the animation down - I spent most of my time lighting yesterday - it’s good that I’m thinking about (especially since I haven’t really done it before) but I can’t afford to do that again!
Yes, I know it’s late! Please forgive me!
Right-click and “Save As…” to download dl_mp_animDaily_003.mov (1.19mb)
Description: Walking down the hallway to the apartment, entering the apartment and seeing the crime scene for the first time.
Arriving at the apartment - I’m pleased with the way it turned out. I like the movement of the camera and I spent quite some time tweaking the lighting to try and make it interesting. This involved pushing Hyde further forward, down the hallway, so that he was under the light more when he came to the relatively long pause where he is just standing, look at the door, before pulling himself together and going inside.
Entering the apartment - This section, like the hallway, can be shot in so many ways; I quite like the way it at the moment. The downward movement (and slight rotation) of the camera actually works pretty well - it was a bit of a fluke because the camera needed to come down a bit for when Hyde walks into the camera. Apologies for the slight feeling of inconsistency as Hyde walks “through the camera”, I think it feels a little odd because I’ve got very few keyframes; with full motion the screen won’t be completely black for nearly as long, but it’s something to bear in mind when it comes to more detailed animation.
Crime scene - This “walking through the camera” shot means I can cut out a couple of horrible camera moves from my previous blocktest. I wanted to do a reveal as is quite common in lots of murder mysteries/thrillers and couldn’t work out how to do it nicely - there isn’t quite as big a build-up of tension perhaps, but maybe that’s something I can add to the entry shot (have Hyde stand and observe the scene a bit longer before moving further into the apartment).
Have to get this blocktest finished by tonight, so I imagine you will be able to see the new blocktest in it’s entirety tomorrow!
Thanks for paying a visit - comments and crits are actively encouraged as always!
Hello! Yes, I know I said I’d post every day, but I had a bit of an inefficient Friday, as I had predicted. The guest lecture from Shelley Page was pretty good, however I don’t think the work she showed us yesterday was quite as interesting as the stuff we saw last year. There was a really cool looking French piece called Renaissance (official French site) though! Anyway, about the animation - yes, finally, at this hour of the morning, I’m starting to be more efficient, so I’ll have something for you tomorrow morning hopefully! Got to try and get this done by the end of tomorrow, it’s already behind schedule!
In theory I’m still on schedule to complete this by the end of Saturday as I whacked out about 800 frames yesterday (and 700 the day before), however most of yesterday’s work was camera rather than character animation work. So I have to be super-efficient today and tomorrow then! That’s a pain because today is going to be split in two by a Friday lecture from Shelley Page of Dreamworks. I went to one in the first year and it was pretty interesting, we get to see lots of pieces of work that she thinks are good, if a little depressing (”no, we tend not to get animators from Bournemouth” :cry:), so I shall definitively be attending again. Oh well, here’s today’s daily:
Right-click and “Save As…” to download dl_mp_animDaily_002.mov (970kb)
Added sections are: dead body up against the wall, extreme close-up of dead body’s eyes and street scene.
I’m not entirely happy with the lighting in the first sequence yet. In the original, the body wasn’t hit by the light at all, but I thought it might help pick out some nicer shapes if I have him at least partially lit. I quite like the way the face is divided by light and dark (ooo, inner meaning!) but the lighting definitely needs tweaking. The problem is that I can’t zoom out further otherwise we’ll see his face, but i can’t be so close you can’t tell what it is (which is possibly just a by-product of having no texturing yet).
The street scene is also something I’m wary of. If cast your minds back to my original blocktest, the street scene was slow to start and then speeds up considerably towards the end. Phill recommended I try and make it really slow and have that juxtaposed with fast paced footsteps and heavy breathing, to create an uneasy feeling. The reliance on sound makes it difficult to tell if this shot will work properly.
That’s all I’ve got for now; there’ll be more later tonight or tomorrow morning. Comments and crits are actively encouraged as always!
You are currently browsing the daniel-lim.co.uk weblog archives for March, 2006.
| Log in |
