The Short Film Storm within was created by myself and Anna McCraith, whilst Anna handled the Character Animation and Compositing, as well as Texturing of characters, I dealt with the 3D Modelling, Rigging, Environment Layout and 3D Effects that you see in the film.
A brief bullet-point of my contributions broken down by scene is below.
- Original Concept
- Major Villains (Conceptualization of the Wolf, Hawk, Snake and Chimera)
- Potential Environment Planning
- Character Modelling
- The Main Character, Alan – Modeled based on original concept art by Anna McCraith
- Fat Rabbit’s – Modeled and adapted from the Model of Alan.
- Modeled based on original concept art by Anna McCraith
- Modeled based on Silhouette Reference.
- Retopology and Model Adjustments to Zbrush Sculpt originally made by Kerry McCormick.
- Model ‘bashed’ using previously listed models.
- Environment Models
- Trees & Shrub Pseudo-3D models based on 2D Textures by Anna McCraith
- Rocks (27 Unique Rocks arranged into 3 Clusters)
- Reeds (Used in original Animatic)
- Attached Thorn Models
- Alan the Main Character Rig & Paint Weight Corrections
- Fat Rabbit Rig & Paint Weight Corrections
- Rig & Paint Weight Corrections
- Rig & Paint Weight Corrections
- Rig & Paint Weight Corrections
- Rig & Paint Weight Corrections
- Environment Layout
- Meadow Scene
- Stream Scene (Version 1)
- Stream Scene (Version 2)
- Snake Lair Scene
- Snake’s Run Scene
- Hill Scene
- Hill to Ocean Scene
- Log Ocean Scene (Version 1)
- Log Ocean Scene (Version 2)
- Underwater Scene
- Underwater Memory Scene
- Final Step-Off Scene
- Scene Lighting
- Character Mesh Lighting
- Reflective Lights
- Film Colour Sequencing
- Tonal Planning
- 3D Effects
- Grass Effects
- Highlight Smoke
- Snake Eyes
- Wolf Eyes
- Chimera Eyes
- Hawk Wing Trail
- Cloud Render Management
- As-Needed Correctives
- Shot-By-Shot Assessment
- Colour Correction (partial)
- File Fixing
- Sequence Set-up
- Shot Assistance & Planning
- Credit Sequence Planning
- Additional Contributions
- Sourcing and Laison for Music Production
- Sourcing Sound Effects
- Primary Research Gathering
In the plan for our short, the wolf was originally going to be comprised entirely of smoke and electricity, however after looking into how to do that, we quickly decided to simply have it emit some smoke and flame.
It took several attempts before we realized that this was an unfeasible goal to add in the time we had, especially as a team of two with a wide spread of work to do.
The tests initially started simple, adding some smoke and flame to the eyes of the characters to make them appear more threatening.
After a lengthy series of tests and practices, I eventually figured out how to graft the smoke to the geometry of the characters so that it moved with them.
This however, created a different problem.
In shots where the wolf moved, the smoke effect came off blocky and thick, rather than the whispy aesthetic I was going for. When static, it worked fine, aside from a few glitches.
Rendering the successful ones also sometimes created this problem.
This mainly occured with Zync, and was supposedly due to issues with caching the particle effect history, however no matter what tutorials I followed and attempts I made to correct this, I could not get Zync to render the particles properly and without issue.
This was what the above images looked like when static, and as you can see, despite all 3 having identical sky-boxes and light set ups, the results could not be more different when rendered through Zync or a different computer.
This inconsistency then drove me towards Paint Effects, and attempting to create the fire and smoke using that.
Needless to say, it had mixed results.
Generally, the paint effects looked good, but didn’t necessarily fit the aesthetic of the film – and as they rendered fully 2D, without existing transparency – it was hard to get a visualisation pre-composition of how it might look.
As it stood, it’s impossible to tell how this would look in the scene – however the results were more shocking when comped together.
Whilst certain terrifying, and very close to the aesthetic we wanted – the wolf seemed to lose all definition as to what it was when we covered it in flame.
Additionally, the sheer amount of compositing that would need done for this, the hawk and the snake proved to be quite excessive.
Given these factors, we decided to keep the smoke and flame effects to a minimum, using them sparsely to accentuate danger through after effects.
This tutorial was interesting, and in an ideal world, we’d have liked our film to feature effects like this on the characters. – However, as you can see, even creating a simple static effect took time – and replicating this through dozens of shots on a moving creature proved… troublesome at best.
This was the idea of how we could go about doing the shot in which the wolf disintegrates into smoke, using a combination of render layers and effects to have it disappear.
As can be seen here, most of the smoke effects we ended up using were simple ambient pieces, with the occasional piece of particle effects around the eyes and mouth, like shown at the beginning of this post.
The primary 3 effects through the short film are the lightning, the grass (and other plants) and the water.
The Lightning went through several iterations before settling on the final version, however most of them ended up looking clunky or throwing off the lighting in the scene.
Ironically, the simplest execution ended up working the best, with the standard Maya lightning effect working very well.
The lightning had to be animated across the scene, to follow the wolf’s position and timed to create the right aesthetic.
With the exposure settings, timing and a basic locator rig that let me drive keys along so that the lightning stretched and ‘vibrated’ more the larger distance it moved.
This tutorial was helpful for understanding all the features and parts of the lightning tool that I could utilize, and how to sculpt the effect I wanted.
This was how the lightning looked when comped into the scene – I would have ideally liked to be able to spend a little bit more time thinning it out and making it more defined, possibly even adding hard outlines to it, but there were other parts that needed addressed with higher priority.
Like I’d mentioned previously, I had made efforts to construct the grass out of Xgen, however it wasnt until recently, that I realized that combining the paint effects tool with polyscatter would achieve the same result with greater ease and effect.
As can be seen here, the Xgen based scenes ran very intensive and didn’t produce an ideal result, especially as any animation they would have needed to be manually implemented.
However, with paint effects, I could create something quickly and simply that was animated by default and resembled the kind of painterly grass we wanted.
By creating multiple different ‘scales’ of the same grass type, I could scatter it over multiple surfaces without creating obvious gaps and maintaining the ‘structural’ functionality of the scenes.
Something that Xgen failed to do, as it often caused crashes due to the intensity it functioned at.
Polyscatter allowed me to instance the grass I made along any environment surface, though it’s downside was that as the instances didn’t generate their own effects, but rather pulled from the parent, they wouldn’t interact with any other geometry in the scene.
This meant that for any shots where the rabbit needed to run through grass, the grass immediately around him had to be removed and manually painted in with collision effects.
As can be seen in this scene, Polyscatter could also be used to add things like thorns across the brambles or to quickly spread clusters of rocks across the ground, all I had to do was import geometry I created and adjust the settings to find a result I wanted.
This was the tutorial I used to learn about the settings and options available to me when creating the grass.
Like in the above video, many of the scenes had multiple sets of instances, that slowed them down immensely, so it was important to use some of them sparingly, and be carefully about the amount of detail we included in the environment.
Not just for the purposes of maintaining scene fidelity, but also to ensure the aesthetic we wanted remained intact.
Water was one of the trickier parts to grasp, and we had many options available, each with their own pro’s and cons to deal with.
Initially, I attempted some bifrost pieces of water to see how they would look, however I found it was too finicky and produced results that were too photo-real for our purposes.
When at a lower setting, like above, Bifrost tended to generate an almost plastic appearance, regardless of lighting adjustments.
This improved the more detail you began to add. however at this stage it was too much like an ocean to be rendered effectively in something like the stream.
This would be great for the ocean, but it took a while to achieve and was very finnicky about how it rendered out.
However, I found that the best result could be achieved using Maya’s built in 2D water texture.
Adjusting velocity and ripple speed let me create a single preset that I could scale through a scene to have the water increase in intensity and height, that way it didn’t need manually adjusted between shots.
The water here had a much more painterly aesthetic that I really liked, and fit well with the rest of the Short, it was also incredibly easy to manage once it was set up – though like mentioned before, it’s pacing was sometimes hard to wrangle.
By placing the water texture on a transparent plane, with another layer on a very mild ramp shader behind it, I could manipulate the appearance of depth.
In the example of the log ocean above – the water layer is highlighted and was set up with a high level of catclark subdivison to generate a nice deformation, whilst the layer below had a displacement map and a ramp shader to give the appearance of terrain under the water, and to back cast a nicer effect onto the water surface.
This was ramped down for scenes where the water needed to be calmer, like the stream or scene 11.
The process of rendering the short film was a drawn out, difficult and often troublesome process, that seemed at times to rely more on luck than any effort of planning.
As Anna finished the animation in shots, I took them, set up the environment scene references and textures – and began the process of lighting and test rendering.
Most test renders turned out something like this however, until eventually I hit the jackpot.
Generally the animation process took some time, which afforded me the opportunity to work more on lighting and set up.
As can be seen in the plan below, some scenes required a constant shifting of assets and the occasionally keyed set of lighting. This was particularly prevalent in the shots with lightning, which had to not only be created, but rendered on a separate layer whilst still having it’s accompanying lights timed to work in tandem with it.
Zync was a blessing and a curse when it came to this, we had to use multiple accounts, amassing approximately £3000 worth of rendering, if this had been something we paid for out of pocket, this film would never have gotten completed. Fortunately, utilizing the free trials available to us and rendering layers on efficient settings, we were able to cut down dramatically the cost and time needed to render.
Shots frequently errored out during the rendering process, or came back with problems that seemed to be temperamental in the fixing. Often, simply closing and opening a scene and re-starting the render fixed it, or simply hiding a cube somewhere in the scene.
This has convinced me that Cloud Renderers, while vital, clearly have some anger issues.
Renders often came out like this, with the light not casting as defined shadows or appearing too dark, whilst this could be corrected in post, it was much easier to compensate for it directly through lighting.
This meant having some layers ‘overlit’ whilst others had to be underlit. This is mostly down to how Zync dealt with the colour output management.
Most of the testing was sporadic frames or short play-outs of how the lighting moved in the scenes.
Generally it worked out without problems once we got it set up, though occasionally we had some issues.
- Floating Geometry
- Non-rendering environment pieces
- No loaded Animation
Generally, these were all easily fixed using alembic caching and baking the animation before exporting.
Most rendered shots needed minor colour correction, basically ensuring they all had similar tones and palettes per scene.
You can see in this pre-correction set of renders, that each character, whilst rendered with the same set of lights, came out with clear outlines and slight pops.
The Wolf had to be rendered brighter, with a purple texture to allow him to be more easily comped and have the effects that drift with him added.
In addition to the Zync system of rendering, we also set up a number of Mac’s rendering independently, and thanks to the referencing system established, this went by fairly quickly, as a scene could be loaded instantly on any mac, with the render setting presets imported directly.
However, the Mac rendered out noticeably lighter than Zync with the same settings, which resulted in an interesting battle when it came to color correction.
The Hawk Rig originally functioned well enough, but it had some scaling issues that made some of the wing flex animation appear wonky, and the controls weren’t exactly clear in their purpose.
So I rerigged it to fix the scaling problem and then created a more clear layout for the controls, as well as adding some stretch capability to the wings so that they could enhance the silhouette.
The Rig is mostly fairly self-explanatory, with basic wing controls to allow the different parts to bend and twist, as the Hawk only appears briefly in the film and is mostly moving in dives or glides, he didn’t need excessively complex components.
This tutorial by Mike Hermes formed the basis for what I chose to do, though I simplified it immensely.
As you can see this early snake rig was fairly basic, lacking alot of features and smoothness it needed. It was generally difficult to animate with and had many skin weighting issues.
To clean this up, I produced a couple different set ups, some that automated coiling for one scene, another that allowed each piece to be moved independently, and eventually combining these into a single Rig.
This was the Rig we’d eventually use, with all features combined – which you can see in action both in the final film and in this clip below.
The Rig had some issues, it’s bending wasn’t as smooth as I’d like, and it didn’t have as tight a coil as I’d have hoped for, but it functioned well in the situation.
The Rig suffered some weighting problems due to the thorns, which threw off the initial weight spread and had to be manually painted to the rig.
This didn’t prove to be an enormous issue – though I would like to have redesigned the rig to not include the thorns, though at this point, the blendshapes had been made and for the amount of impact it would have had – redoing them was not worth the effort.
This tutorial form the basis for the above Rig, although the exact interpretation of this lead to some issues, namely that it couldn’t form the shapes we needed it to – hence the more complex adaptation in the film.
This Rig was the replicated version of the one above – I had simplified the body which meant it needed reskinning – however during testing, it didn’t move quite the way it would need to, so we decided to simply fix minor issues and adjust the original instead.
Having completed the project, I can now reflect on the Rabbits Rig, what I learned, achieved and would improve if I were to create it again today.
Initially the Rig had alot of problems, and some of them still persist at completion.
Primarily the Rabbit can’t stand up on it’s hind legs without some major deformation issues – however research showed that the only solution to this was to have two seperate rabbit models and rigs, one built for bipedal motion and another for quadrupedal.
Seeing as the rabbit had only two scenes where he needed to stand up, this felt like an excessive diversion at the time.
Paint Weights and IK issues also appeared prevalent, and fortunately Alec was able to help correct many of those, which I could then apply to the Fat Rabbit in turn.
One of the interesting parts and adjustments made to the rabbits Rig was the functionality for chest and joint swell, which was useful for shots like the one above, where his head had to crane back and lift, which would normally cause excessive deformation and texture drag.
This was fixed with a combination of a driven joint and a blendshape that would pull in parts of the lower belly and push out the neck and chest to make it look like his body was correctly shaped from odd angles.
Constructing the Fat Rabbit rig mostly involved the transferring of attributes between the Main Character Rabbit and the Additional Rabbits.
This was fairly simply done after uncovering a couple helpful tutorials.
Transferring UV’s made painting the skin weights much easier, and meant that we could also transfer the skin weights on the Rabbit itself to it’s chubby sibling, making cleanup more effiecent.
This video was less useful but clued me on to the actual transfer of the rig to a new model and came in useful when I would add a feature to one rabbit that I wanted to replicate on the other, saving a lot of set up time and leaving only the bug fixing to do.
Autodesks own tutorial on Corrective Blendshapes was more than sufficient for the small amount we needed to do. Mostly consisting of wrist and belly inflation to compensate for the stretch and allowing the rabbit to move with a bit more weight.
Over the past few months I’ve followed alot of tutorials, below is a rather extensive list of the ones I followed.
I used this tutorial primarily to get a stronger understanding of how to skin the joints of the character, and to try and find a solution to the ‘floating eyes’ problem I encountered.
Floating Eyes: The Eye, Teeth and Tongue Geometry of the Snake frequently detached itself from the rig and moved through space parallel to the rig.
When Modelling the Rabbit and Snake, and early attempts at the wolf – I would create their general shape in Maya then port over to Zbrush for fine detailing – I hadn’t touched Zbrush for some time and so I used this tutorial and the one below to refresh my memory on how to use it.
This Tutorial and the two below are fairly obvious, Mainly trying to find different ways of modelling the eyes to make them look realistic – Anna was fairly strong in her view that the eyes needed to be realistic to make the characters feel alive, and so I wanted a range of options.
This particular tutorial was quite handing for working out how to get eye-reflection, however it didn’t work as well in Arnold as it did in the early test renders on maya software.
The mouth cavity tutorials were mostly to just get an eye for the topology and working out how the blend-shapes would deform.
This particular tutorial was mostly to try and get an idea for how I wanted the mouth to deform with teeth and still appear threatening, though I ended up going for the more realistic approach with the teeth and have them contained within the mouth instead of protruding out.
I love the look of this rig and some of the more interesting components of it, though it was more complex than what we needed for the short.
This tutorial was particular useful for grasping how the leg set up should work, with the full body crouching and hierarchy structure of the controls.
This and the tutorial below, primarily served to get a sense of how the legs of the wolf and chimera should bend – initially the chimera was a completely different creature from the wolf, exhibiting similarities and features, but eventually it became easier to simply adapt the wolf itself for narrative reasons.
Though we had no human characters, these were mostly to help get a sense of IK set-ups and refresh my memory of some of the tools available to me while rigging.
Similarly, this tutorial was to help me quickly grasp joint hierarchy and remind myself of the different settings and options available when setting up skin weights.
Whilst this tutorial was helpful, the features it showed from DM were never used, as I couldn’t actually source a free functional version of it for 2018, though the 2017 version was interesting to play around with, if a bit clumsy looking.
Initially, the facial animation was going to be done with controls, however blendshapes were a stronger and more flexible choice for what we wanted.
This whole section was part of me trying to work out why I couldn’t get my IK’s working the way I wanted them too. It eventually took a tutorial Alec sent me from Lynda.com for me to figure out the solution.
This tutorial was less about the IK and more about the arrangement for getting the wolf leg motion exactly how we wanted.
Quadrupeds are tough and make an interesting challenge.
This first foray into Paint Effects seemed to convoluted for me to want to use, however I would later come back to it when looking at how to craft the grass for our scenes.
I loved working with XGEN and MASH grass, however it proved too scene intensive, eventually preventing my laptop from opening any scenes at all – as such, the environments had to be rebuilt and a less strenuous method found.
This was used to help guide me on how I was going to create the busier stream section, and while this tutorial builds a much more complex environment – we wanted to keep ours simpler to help maintain focus on the characters.
These tutorials were just research into different options for creating grass I had, and hopefully in creating fur on the characters – though that would eventually get scrapped for myriad reasons.
Like I mentioned earlier, I initially looked at attaching fur to the Rabbits and Wolf, however similar to with the grass overloading my environments – this made animating and fixing rigging issues very difficult, and it became more hassle than it was worth to actually graft the fur to the characters.
Most of these videos highlight the things I either got stuck on or needed to refresh myself on how to go about enacting. They were mostly helpful, with a few duds here and there depending on many different factors at the time.
Initially our colourboard would have used real wolves, but real wolves lacked the nightmarish colourscheme we wanted, and so we switched to concept art and abstract-ism.
Fortunately, we not had lots of wolf images to draw from, a handful of which included nice clear images of snarling wolves. As this would be the primary expression for the wolf in our animation, I figured it would serve well to post them.
The first thing that is apparent is that the nose and the teeth become the most prominent parts, with a large mane of rough and the eyes becoming more like small beacons against the furrowed brow. A good thing to keep in mind when it comes to designing and modelling the creature.