ECopeland MA 3D Computer Animation

After around three days of day-and-night attempting to figure out why the ground plane and geo were not moving as expected with the camera, I was growing frustrated. I’d even resorted to moving the scene around with the camera, effectively messily counter-animating, but after two render attempts I had to accept that this was not only a waste of time, but also would never achieve the desired results.

I’d attempted to parent the scene to the camera, and I’d looked into some of my old tracked shots to figure out what I’d done wrong, but could find no answers. Growing weary of searching for a solution, I forced myself to make a concrete list brainstorming possibilities.

I decided, overtired and unmotivated, that the easiest thing to do for the time being would be to look into 3DE’s orientation view and see if there was a mismatch in how the camera interacted with the scene in Maya vs 3DE, although I strongly doubted that the issue I was experiencing was related to the track and suspected more that it was due to the way I’d set up the scene or my hierarchy, though I couldn’t figure out the exact root of the issue. Nevertheless, I took a break from fruitless counteranimating to delve back into 3DE.

As far as I could tell, it looked completely fine. The camera was moving towards the stationary points as I’d expect it to, as I wished it would, in the Maya scene. Interacting with the points as it should with the geo.

That’s when I had a big moment of realization.

When I set up my scene, I had begun building my ground plane without snapping it to a point. I must have forgotten to do this, or else I’d assumed that if it looks correct in the camera view the camera will still orient correctly around it, but it suddenly made sense to me that although the scene geo looked right in the first frame, it could still be an incorrect distance to interact like the points were- for example, I could’ve placed it much too far away and simply made it large enough to look correct without actually being correct, thus when the camera moves closer, it is barely noticeable because the geo is much further away than the points.

On a hunch I hid my animation, deleted my ground plane and building geo and reconstructed it, snapped to a point in the scene this time.

When I hit play, it was a miracle: the camera moved around the scene correctly, the ground plane stayed kissing the edge of that building perfectly.

I reconstructed my building geo, making sure to extrude the edges of my ground plane this time to build the background set and reduce the mesh to as little faces as possible, my logic being that if I know the ground plane works there’s no point giving myself a second problem with complex geo elsewhere/ interacting on top of it.

And, blessedly, the geo and ground plane both moved correctly- that is to say, the camera moved around them correctly- in the scene.

The night had stretched on into the early morning hours, but I resolved to set my scene back up and put a test render into the farm to see if my plan had worked in the morning.

I resized my scene and moved Dana back into place.

She’s now surrounded by points. As I had mentioned in my notes, I’d been experiencing some crashing in Arnold Renderview, so I wasn’t sure whether it looked how I wanted. No matter, I’d put the render through and if it worked, I could re-adjust one last time before my final render.

The Next Day

And, after several days of frustration, it had worked. The scene was finally set up as it should be.

The excitement and relief I felt was immeasurable, and I liked the lighting setup even more than I did originally. Tasting an end to this leg of the project, and ready to move on to the next, I took just a couple notes for tweaking the scene to its final glory:

It took me all of thirty minutes to make these changes. Ready to embrace the product of my hard work, I changed the file name to Final_PinkDana render (or something of the sort) and chucked it into the farm.

However, so close to the end I could taste it, fortune did not smile on me, as I encountered an error with the farm: Deadline was marking jobs as complete, yet not saving the files to their output destination, despite the paths being correct and identical to the setup from last night. After several trial-and-error attempts to see if there was anything I could do on my end, I reached out to Luke for help.

Expect a glorious final render in a separate post later today.

In the span of a night, stretching into the early morning, I was able to track the footage I’d spent days carefully choosing (I had originally chosen a different shot, actually, but it involved a stationary camera, and this presented a problem with determining the depth of the scene; not only did I not want to involve a nodal camera but it also fails to display my tracking skills well in a showreel). Fortunately, this shot, which I’m referring to as Pink_Wall, tracked smoothly and easily.

The shot itself is a one-of-a-kind on Pexels, with camera motion and a ground plane in the frame while still depicting no actual humans. I’d spent days perusing Youtube and alternate stock footage sites, frustrated with the utter lack of stock footage that appropriately depicted a usable background without actors in the frame.

After tracking the footage, I glided through setting up the Nuke composition and snapping the 3DE scene to the ground plane. I was ready in no time to build the set, which is one of my absolute favorite things to do.

This is the composition I created, and I’m almost as proud of it as I am of my Term 1 set from December. [Artist Moment] That one was great work because of the light setup framing the character and highlighting the blood splatter dramatically, but this one is good work because of the way the color patterns bounce the eye around the scene. I’ve chosen green and yellow leaves to draw the eye to the bus stop at the right of the screen as well as the yellow of Dana’s shirt, and the funky festival posters bring attention back the the retro-neon pattern on Dana’s trousers, though shrouded in shadow they don’t steal the stage. The color of the arm in the street art on the left draws the eye to the tan of the tree bark on the right, framing Dana and her action. I’ve tried to include props that place the scene in a time period relatable to the viewer without giving the city a specific location and posters that keep the atmosphere vibrant and cheerful.

I chose to give Dana a bright and stunning outfit to match the personality I will pair her with as well as to help her stand out in a high-contrast scene. It was a good choice as well to give her bright gold earrings, as it adds that extra touch to her face and personality.

After setting up the scene, this was how my shot looked:

All set for a first test render.

However, deep into 3 AM, I discovered that my scene setup still needed a lot of work.

The ground plane, applied with a shadowmatte, is moving incorrectly, and appears to be rendering incorrectly as well, which will present a problem as I fix it in the way her feet interact with ground.

The work I have to do before this is done mainly involves fixing the shadowmatte groundplane. I also intend to animate gentle wind through the leaves overhead.

This is the test render of Clara, which I created for the same reasons as the Dana test render from yesterday. With hyper-realism it’s harder for me to tell whether it’s perfect, but I can no longer spot glaring mistakes.

I watched this split-screen with the Dana render and the effect is really, really cool. It’s really exciting to finally begin to visualize what this project will look like when it’s completed! I’m starting to already get a sense of pride in this work and it’s amazing.

This is a test render of my Dana animation to check that it all works correctly on the farm and won’t present an issue with crashing for days on end. It also helps me get one more look at my work and spot any last changes.

The hands definitely snap back up a little bit too fast before the “mm-hmm.”

Next step (aside from fixing this) is tracking my footage and putting the scene & Nuke composition together. This is usually easier said than done. I’m also so proud of this animation that I can’t even decide on footage to track for the background- I’m thinking maybe just blurry city skyline + add in some 3D assets in the foreground. Normally I choose my footage before I decide what the animation will be, but 3DE has been down on VMWare for some time and it does not work on my home computer. As it’s still down, I’ll resume my work on Clara before I move on to my tracking for Dana. If I finish both at the same time it may be a good thing- to take a bit of a break from animation before I move on to either the realism segment of the cowboy project or start my third and final pairing.

EDIT- minor, of course, but I may want to update the texture on that phone so it looks like she’s actually on a call

Clara is the hyper-realistic counterpart to Dana. Beginning work on Clara was cathartic, because I finally got to see the pieces of my FMP idea coming together. It’s incredible to already be able to compare and contrast the two animation styles and imagine the finished product, it’s really given me the motivation to push through with it after the frustration that was getting stuck on the cowboy WIP.

The Clara rig itself has not given me technical trouble, other than some irritating joint weights I’ve had to repaint and strange teeth duplicates that were constrained incorrectly. Thanks to my new (yet still quite young) experience in rigging, I was able to identify and solve these issues on my own pretty swiftly.

The only strangeness in animating the Clara rig comes from the fact that I’ve now spent over a month going into work and putting my energy into creating the most cartoonish, rubbery, absurd and zany work that I can. Putting myself in the mind of a child playing a video game with a very short attention span and a juvenile sense of humor. This has translated well for Dana, but as I work on Clara, I find myself wondering continually whether my animation is “enough”. It feels like almost nothing. A small, subtle hip movement every 50 frames. Very precise neck angle. Dana feels like opening a sandbox and just seeing what I can create, Clara feels like tracing an outline carefully.

That said, I am proud of my Clara work, and it’s fun to compare them.

As it stands right now, I’d have to say I personally prefer the cartoonish animation style, but this may be a premature call. I haven’t completed a finished render for any project to compare the two, and that realistic cowboy does have a lot of intrigue.

Anyway, I plan to have a cartoon-realistic shot for a total of 3 pairings, each with their own tracked shot. If worst comes to worst and I can only do 2 pairings, I’ll still have 4 tracked shots and 4 animations, which isn’t bad. I do need to make sure I have time to get my thesis done as well, which has kind of been placed on the back burner for the time being, as I know that I can write quickly and effectively even in a time crunch while I simply cannot cut corners or “just work faster” on animation in any way shape or form without taking huge sacrifices.

As opposed to the Cowboy WIP, I’m very, very proud of Dana. She’s already my best work. Dana represents the stylistic and technical skills I’ve been picking up at my internship- learning to be comfortable with extreme squashing and stretching as well as implementing cartoonish effects in a way that’s both entertaining as well as continuing to make sense.

I love the Dana rig. I discovered it when reviewing portfolios while going through applications to hire onto my animation team last week, an animator had used Dana in a facial animation expressiveness test. I recognized her as being stylistically similar to the David rig, and when I found the creator I downloaded both her and Amy. Sure enough, Dana is a dream to work with. Every control allows deformation and has full options for stickiness, roll, etc. The fingers and face are rigged just perfectly. Although there isn’t phoneme availability, the quick expressions are wonderful, and using Studio Library I was able to build up a phoneme database for the lip sync within less than an hour.

I also did some re-texturing, of course.

But what I was most excited for and what I love most about the Dana rig is the numerous hair controls. as you can see from my work so far, I love animating Dana’s hair.

The character I’ve been animating in my internship is a raccoon with great ear controls, and one of my favorite tasks is to “add ear floppage”. I’m insistent that he needs to have “jelly ears”. This, when applied to Dana, is the jelly hair effect. I’ve in fact made Dana quite gelatinous in general.

I have very little setbacks to report here. I’ve been working with Dana for a little over a week and have almost completely finished animating her, running into absolutely zero issues. My animation workflow is just as it should be- I’m able to focus on questioning my stylistic elements without worrying too much about tech-related difficulties.

As far as questioning stylistic elements, I guess I will say that I may have gotten somewhat carried away during that “aww”. I know it’s a bit extreme, but I love that rubbery bounce so much, it would hurt to remove. Perhaps in juxtaposition with Carla it’ll be more acceptable. After all, this half of the project is not meant to explore realism.

https://vimeo.com/597486076

The Cowboy WIP has been dwelling uninvited in the back of my mind. In contrast to my Dana and Clara work, it has not come easy to me at all, and it’s apparent in the much lower quality of work comparatively to the standard I’ve been recently holding myself to. As it stands right now, it doesn’t deserve a place in my showreel, and I don’t want to turn in work that I’m not proud of. I don’t know why I thought it would be easy, perhaps because it’s a “simple walk cycle”, but it’s been fraught with constant frustration, not entirely due to the fact that it involves both a quadruped and a bipedal character moving simultaneously.

Before I begin, I will just remind the reader of my intention here. This animation is the exaggerated, cartoonish counterpart to what will be a hyper-realistic animation featuring a much more intimidating character.

Anyway, my initial issue with the cartoon cowboy shot was a question of how to parent the hands. I intended to have the cowboy’s left hand stay placed upon his belt- as I had planned in my reference footage- but, still having yet to fully embrace the functions of working in FK, I wasn’t sure how to accomplish this, and was frustrated by the mesh deforming when I placed the hands in IK, due to the fact that both were parented to moving objects (the hip and the horse). Luke helped me find the answer to this.

In the same vein as parenting the hands, I had taken the rigged hat from another character and given it to this one, and I struggled to figure out a way to parent it to the head and animate it, yet allow him to take it off later. I tried baking the animation and working against the parent constraint, but ultimately my solution was to duplicate the mesh and toggle visibility- there are actually two hats. I’m not altogether happy with this, though, because it’s messy and because I would prefer to be able to animate the second hat as well.

In addition, I can barely even begin to get into the struggle that was trying to get this onto the render farm about two weeks ago. I’m not sure why, but the frames 34, 148, and 181 (oh my god I have them memorized) repeatedly failed to render correctly on the farm, rendering the perspective camera rather than the tracked footage. I tried rendering the EXRs manually and replacing them, I tried duplicating the camera and rendering the specific frames with that one instead, I tried parenting a separate camera. The most baffling part of it was when I finally got those frames correct and instead 33, 147, and 180 failed. As I pushed through this, I also worked to get my Nuke composition done right, having only used it a few times (though more and more frequently as of late). In the end, though, it wasn’t Nuke that was giving me trouble, even with the complicated file pathing that fills me with anxiety, it was just Maya and its inexplicable failure to render those frames. Shamefully, my solution was finally just to duplicate the frames before each failed one and replace them. It’s not really a solution so much as it is avoiding the issue, but the animation itself needs work and thus is going to need to be re-rendered at some point anyway.

Another huge, overarching struggle, the biggest one you might say, is twofold. Part of this is animating the horse in a way that looks real while of course not having access to reference footage the specifically works for what I’ve planned. I’m working mainly from a muscular diagram, which can only do so much for me. After asking for feedback from multiple animators I’ve gotten the horse animation to a place that isn’t abysmal, but it could be better. The second issue is being unsure where to draw the line in exaggerating the cowboy movement. What is “dramatic” and what just reads as a mistake? I feel that because I’ve been staring at it for so long, it’s hard for me to tell.

I will get help from Luke on this one shortly. It is embarrassing to show work that isn’t good, even if that’s the point of asking for help on it.