The Making of Collectible
Collectible was a year long labor of love for everyone involved. If you’re a member of StoveGhost Studio, I hope this books helps you think back fondly on our semester. If you’re an ETC student, we all hope this book will inspire you to succeed at your project and help you learn from our mistakes.
We had a hell of a semester, packed with excitement, stress, failures, and success. A film is never complete, and there are always things that could have gone better, but we’re happy with the work we’ve done, we did a lot of things right, and we’re excited to share what we’ve learned.
Project Beginnings
The start of StoveGhost Studio was a conversation between Steve Geist and Freddie Sulit. It went something like this:
FREDDIE
Man, an animated short would be a fun project.
STEVE
Yeah, yeah it would!
FREDDIE
I should make a spreadsheet!
STEVE
This lunch has been both delicious and a defining moment for our futures.
The Team
StoveGhost Studio’s membership fluctuated a lot leading up to the pitch. Our initial development team included:
Andrew Gartner
Michael Capristo
Michael Honeck
Steve Geist
Kim Kiser
Tom Luong
Freddie Sulit
Franz Mendonsa
Along the way we lost a few members to internships, co-ops, and other projects. Freddie picked up a production gig at Pixar, Franz headed for the sunny beaches of Schell Games, and Tom flexed his producer muscles on the Pandamonium! project. We picked up Patrick Jalbert as our animator. Thus, the final team became:
Andrew Gartner – Technical Artist
Michael Capristo – Writer, Sound Designer
Michael Honeck – Producer, Editor, Compositor
Steve Geist – Storyboard Artist, Texture Artist, Lighter
Kim Kiser – Art Lead, Texture Artist, Lighter
Patrick Jalbert – Animator
Not a typical ETC project
Animation is always tougher to sell at the ETC than other types of projects. StoveGhost was unique not only because our focus was animation and storytelling, but also because we were avoiding pushing any technical boundaries. We felt that for our project to succeed we didn’t need to innovate, we needed to successfully entertain and execute our vision.
We were told from the beginning that we’d be in for an up hill battle for approval, so we got down to plotting our path to success.
Inspiration
With a team and a goal in mind, we began looking at what kind of animation and stories excited us. We spent the summer of 2010 scouring the Internet for shorts we found touching, funny, and visually exciting.
Brainstorming
To brainstorm, we all took stabs at writing up one-page story summaries called beat sheets. These verbal sketches let us come up with gags and plot points quickly and without much visual development. Because we didn’t get too attached to any one drawing or idea, we could create 6 characters in 6 different situations, judge their appeal, and dismiss them in a matter of a few hours. When we finally settled on a story, we were (fairly) sure that our basic backbones of character and plot were appealing and solid.
Pitching
To pitch successfully at the ETC, you must be part salesperson, part politician, and part clairvoyant. The faculty want to see an idea well presented by an excited, capable team that knows what they’re talking about and what they’re getting into. As if that weren’t difficult enough, you’ve got to contend with the successes and failures of past similar projects, and the personal feelings and opinions of the faculty. We came very close to losing the pitch, but the faculty saw the potential for success and gave us the push we needed to become a viable project.
Before we went into pitches we had a basic story draft about an old woman who buys things until it kills her. It was fairly slapstick and had a fart joke in place of Cutigan’s First Cutigan (Heaven Scent, ha!). Mk was our faculty champion. In round 1 we pitched the basic idea: we are a handsome group of talented folks want to make an animated short. The faculty’s response? “Ok, we’re interested, but we want to see a better story, and let’s have some art.”
Round 2 did not go as smoothly as Round 1. We attempted to develop our story further while going to the faculty for feedback, while developing the art, while also working on our various semester projects. We went into Round 2 with a less sophomoric script, but it still wasn’t tight, the ending really didn’t land, it was a clear case of story by committee. Our art looked generic and uninspired. We were all a bit frustrated because we felt that we really could do it, if given the time to work exclusively on our pitch.
This could have been the end of StoveGhost. After all of our hard work and extra hours, we were all ready for the relief that would come with news of either success or failure. Instead, the faculty made us an offer we weren’t ready for…a third, final death round. The deal was this: the faculty wanted to see 3 distinct visual styles along with a written explanation of each style. We were then to choose the best style and argue why we felt it was ideal for our project.
We developed our styles and reviewed them with Kevin Allen, a NJ based Set Designer and Art Director. He helped us think through our process and come to the best style for our time frame and story. We sent our findings to the faculty and a few days on pins-and-needles later, we were approved and up and running!
Finding our Style
American Regionalism & Handmade
American Regionalism was inspired by the paintings of Edward Hopper and Thomas Hart Benton. Elements of the style include: visible brushstrokes, exaggerated and elongated human forms, soft lighting from practical sources, and painterly textures. This style would serve as a contrast to the cold isolation of our story. Lenora, our main character, is a grandma, traditionally associated with the same warmth that these artists are known for, who becomes a crazed, pitiful character whose disheveled look ultimately lends itself to Benton’s distorted, bottom heavy portrayal of human form. This style especially contrasts with the Cutagins who can appear soft and painterly on TV, but cold and dead eyed once unboxed.
The Handmade look was inspired by claymation and stop motion animation, such as work from Tim Burton and Aardman Animation. Elements of this style include: oversized textures, exaggerated and cartoony proportions, asymmetrical objects, soft lighting, and a feeling of claustrophobia due to the small scale and large textures of the world. This style is associated with loving attention to detail, which compliments Lenora’s love for her grandson’s gift, while contrasting against the mass produced Cutagins.
We chose not to pursue these styles because the 3rd style we explored did the most to enhance the ideas behind our story. The Handmade look in particular was worrisome due to the choppy animation style – it had the potential to look like a bad student animation.
Cartoon Modern
Cartoon Modern is the animation style of the 1950s and early 60s. This
was a period of evolution in animation, and is in many ways a departure from the original, soft, “illusion of life” approach pioneered by Walt Disney Studios in the 1930s and 40s. This is a more graphic and abstract approach, influenced by the modern art movements of the time including Cubism, Abstract Expressionism, and Surrealism. Elements of the style include: ambiguous depth, flat panels of color, heavy outline, hard lighting, importance to story determines an object’s level of detail in a scene, lighting has a strong influence on an object’s local color. The flat panels of color create dramatic lighting through the long, hard light/shadow boundaries across scenes. Reducing the detail level of the backgrounds can help to emphasize our character’s increasing isolation. Because environment color can change drastically in this style, the increasing darkness of our script can dramatically affect the color palette over the course of the short. The simple, graphical outline of background elements compliments an environment that is packed with faceless, meaningless knickknacks. We were excited about this style because it is a translation from 2-dimensional visuals into a 3d world, which we had never seen done before. The look provides many tools to help push the emotional impact of our story through lighting, color and detail.
Visual Development
Designing the World
Now we had to design the world in the Cartoon Modern style. We had a bunch of visual development sketches from the 3rd round of pitch, so we built off of them. Unlike live-action film, every environment and prop has to be designed and created inside the computer, and remain consistent with our style. We did a ton of research: we watched cartoons, looked at still frames, concept art, and designs from cartoons of this era from studios such as Warner Brothers, Hanna-Barbera, and UPA. Even employees at Disney were allowed to experiment with Cartoon Modern on some shorts. For example: Toot Whistle Plunk and Boom headed by Ward Kimball. We were particularly inspired by Disney’s feature film 101 Dalmatians. The book Cartoon Modern, by Amid Amidi, which documents this period was also helpful. We also researched several important figures such as:
Ed Benedict, who shaped the early Hanna-Barbera design style with Yogi Bear, The Huckleberry Hound Show, and The Flintstones;
John Hubley, the director and creative head of UPA at its peak;
Ward Kimball, who created most of Disney’s graphical shorts;
Ken Anderson, 101 Dalmatians art director and production designer;
Walt Peregoy, color stylist for 101 Dalmatians who was responsible for the large planes of flat color that only loosely followed the line drawings;
and Mary Blair, who created concept art for 1950s Disney films.
Prop Development
Every prop needed a model sheet, and we pulled from a lot of reference images to make sure everything stayed in style. After the model sheet was completed, it went to modeling and texturing. Because we had about 100 objects to model, we had many classmates offer to help. They did a few models apiece over Christmas break, then we finished modeling internally during the first quarter of the semester.
Developing our Render Style
We knew we wanted a dramatic color shift as the short intensifies, so we looked again at 101 Dalmatians to develop several color stages. The short begins with subtle blues and greens and ends with a fiery palette of reds, oranges and yellows as Lenora descends into a metaphorical hell. To create the environment we started with a bunch of different sketches of various parts of the house, and then created the final layouts and top down view to get the environment ready for modeling.
Lenora’s Design
We wanted to make Lenora feel like a typical grandmother. She had to feel as appealing as possible because we did not have a lot of time to develop her character. When we first conceived of the character, we looked at how older women were portrayed in animation. we also looked at famous older women. Betty White became a major influence on the early designs. During the early phases of the design process, we created a character attribute sheet. This helped us determine how Lenora should act and feel about her obsession. When we started trying to figure out our overall visual style, Lenora’s design played a major part in the decision making process. We looked at how we would portray a grandmother in each of these styles, and what these interpretations said about the character. Our first pass at Lenora’s design in the cartoon modern style made Lenora feel too “high status.” We wanted there to be more tension when she begins spending all of her money on Cutigans. We also got a lot of feedback from the faculty that Lenora was not appealing. We spent some extra time on her design and made sure that she was as appealing as possible before we handed her off to modeling. Lenora’s final design featured very round features, a large head, and small extremities. All of these features helped make her more sympathetic, but they also made it a challenge to animate and stage her in the world. While we were looking for an animator to join our team we talked with many first year students. This was good for the team because it sparked a lot of interest in the ETC community. A few modelers asked if they could help out, and we were lucky to have Noah Bench translate Lenora into a model.
Delivery Man Design
We started the delivery man’s design with the same process we used to develop Lenora. We began with a few rough sketches and a character attribute sheet that allowed us to figure out how this character should react and interact with Lenora. Once we landed on the Cartoon Modern style we were able to move quickly with his design. The delivery man’s final lanky proportions stand in stark contrast to Lenora’s childlike design. This contrast is prevalent in Cartoon Modern. Ethan Gagorik created the final model.
Cutigan Design
We had a lot of fun coming up with different Cutagin designs – the final series that made it to the screen included Precious Presidents (baby figurines dressed up as U.S. presidents), a Tortured Artist collection, the Make-um-ups, and a random assortment of others including a loaf of bread, the kitchen sink, and the L. Ron Hubbies and Third Reich. We made sure they looked appealing on the tv, but appalling once they were delivered to Grandma. This was achieved by giving the Cutigans wide, friendly eyes on television, but vacant, beady-eyed expressions upon delivery.
The Story
Initial planning for the project began with no specific story in mind, only a vague theme that everyone seemed to like. The Sisyphean rigors of graduate school inspired us to tell a story that in some way addressed the question “when are we done?” Since none of us was sitting on a great script and since all of us would have to be creatively invested for about a year, we approached the initial story meetings somewhat democratically; each of us would bring in a short synopsis of a story idea and if the team generally liked it, we would develop it further.
We started with an outline for a short about an old woman who collects porcelain miniatures from a shopping network. Her manic accumulation of these collectibles lead to her falling from a high shelf – an idea almost identical to our final product. What did change was the tone and the character arc. The first draft was “punchy”, full of slapstick and as many gag beats as we could cram in. The faculty largely felt that this approach to be unworthy of investment, so we revisited our story, focusing more on character, theme, and what our short can say about obsession and loneliness.
The “testing” and rewriting of our script resulted in a more thoughtful story about an old woman whose loneliness drives her to collect porcelain “Cutigans”. Her delight with ordering these presents for herself gives way to addiction - a joyless accumulation – until her obsession buries her. Adding emotionally honest details to her life pushed this piece from a caricature of a “crazy old lady” towards a glimpse into a more complex, in flux, character.
The Score
Like any film, the music in Collectible was meant to provide an emotional backdrop to what’s seen on screen without dominating the piece. Both the music and instrumentation had to portray the introverted and contained nature of the main character. To accomplish this, the entire soundtrack was performed by a string quartet and piano quintet to give the music a sparse and tight feeling to match the nature of the main character’s lifestyle.
The natural sound of the strings was also intended to clash with the cheesy elevator-type music that takes over during the TV segments, which eventually evolves into its own dark theme when it becomes clear that the TV is shoveling the very things that are wrecking Lenora’s quiet life. The credits roll with the shopping network theme to emphasize the ridiculous and dark nature of the main character’s central conflict.
Storyboards
The storyboarding process began in September of 2010. While we did not have a locked script, we tried to work out the story beats we felt confident about visually. These early boards helped us define the tone of our short and it helped us figure out where our story was and was not working. After many iterations and long, long meetings we finished the storyboarding process at our quarter semester presentation. The storyboards shown on the previous page are the final storyboards that we used to build our animatic. The animatic allowed the team to set a pace for the short and allowed us to start the scoring process.
Transitioning from Boards to Animation- Once the animatic was locked, the team faced an unexpected challenge in translating the timed storyboards to actual animation. Small details that had not been fleshed out in the boards- like Lenora dialing the phone- became obviously unplanned once she was moving. What we expected to be a linear process became very much a back and forth between Steve and Patrick to faithfully translate Lenora’s feelings and actions.
Color Script
We wanted the color of the short to shift with the story’s emotional turns. We created a color script to allow us to visualize how and when this was going to happen. The script not only gave us a bird’s eye view of the changes in color, lighting, and contrast that would occur in our story, but it also gave us a source to pull colors from when we were texturing and lighting our shots.
Animation
Lenora was a silent character, and that meant that her personality would be strongly reflected by her body language. To keep her lively, we had to plan for her to be unusually spry for an old lady. Moreover, Lenora had to become significantly more athletic in scenes where we were pressed for time for her to complete an action.
Because Collectible was a one semester project with one animator, animation was on a tight budget. For many shots we didn’t have time to perform separate blocking and rough animation passes. Instead, shots were ranked in order of their importance to the story, then blocked and roughed in a single pass. The most emphasis was put on posing Lenora clearly, with a secondary emphasis on fine tuning the animation curves to create higher quality animation. The ranking system saved us because it let us go back and re-visit the most important shots for final passes at the end of the semester. The end result was that high-impact moments got the best treatment, and less important shots were saved for last.
One of the things that we didn’t expect on this project was how much animation time complex facial expressions can add. We thought that because our characters were silent, we would avoid time consuming facial animation. Lenora proved us wrong because her facial expressions became just as important as her body posing. Often, this lead to adding another layer to the animation- roughing her poses first, and re-evaluating the scene for a “face pass” afterwards.
The biggest challenge when animating Lenora was her bodily proportions. In the storyboards, Lenora could stretch that subtle amount it would take her to hold something out in front of her, or use a telephone. In animation, mimicking that subtlety proved to be problematic. As a 3D character, her head was so large that she physically couldn’t touch her face with her hands. We ended up using a combination of camera angles, “unique” poses, and some squash&stretch in the model to bridge the gaps between storyboard and final animation.
Production
Our short was huge. Though only 5 minutes long, it contained well over 100 unique props, 2 fully articulated characters, 3 complete texture changes for most assets, and 96 individual shots. With all of the decisions to be made and remembered, errors to be found and corrected, and assets to appear in the right shots with the right textures and lighting, we needed a robust, automated production system to keep track of where we were in the pipeline.
Google spreadsheets became the ideal solution for parsing our mountain of data. We could access documents from anywhere allowing us to work from home and to create production notes or record feedback from anywhere in the building.
The goal of all of our production documents was to allow us to see our progress at a glance and understand if we were on schedule while not hiding the details of individual shots’ and assets’ progress in the pipeline.
Our view-at-a-glance was provided by the Ghost Board- a series of 3 gauges that showed, us on a scale of 0-100%, how our departments were progressing and our short’s total completion. As a part of the pipeline neared 100%, it would come down and a new gauge would appear. This was a great motivator, because a productive work day could move the needle as much as 10% in a department or even for the short in total, making one person’s success celebrated by everyone.
Creating an animated short is a very linear, waterfall-style production. Because of this, Gantt charts were a perfect way to visualize our schedule. Data flowed from our Master Shot List and Asset Tracker to give us a time focused view of our progress. The Gantt chart looked scary some times, and lead to some hard but fruitful discussions about scope, priorities, and scheduling.
Most of our tracking system required very little direct manipulation by the team. The exception was our shot list. The Master Shot List was where every department met to track each shot’s progress through the pipeline. It generated most of the Ghost Board’s gauges, and fed data to the production schedule. The team could watch a shot move from storyboard to compositing and back again as changes were needed. This document was instrumental in highlighting what shots were requiring the most fixes, and what sequences were taking the most time to complete.
Pipeline
Automation
We had scripts and tools that covered a lot of tedious and error prone tasks such as creating geometry caches for animation and copying assets to the renderfarm. These tools helped us take a shot from lighting to the render farm in only a few clicks rather than hours of updating assets by hand. Though usually reliable, occasionally a script would fail and we’d have to process a scene by hand. Our character rigs were also scripted, allowing minor changes to be implemented without having to re-rig an entire character by hand. Ultimately, we were able to automate a lot of tedious of tasks including render-layer setup, leaving artists free to concentrate on art without being responsible for mundane repetitive tasks.
Lattice Tool
The lattice tool allowed us to distort props quickly and easily without having to create a new model. The tool created a 3D grid of control points that could be used to scale and twist geometry without creating new polygons. It was a fast way to achieve the dramatic distortion required at the end of the short.
Muster
The renderfarm was managed by the Muster client. Muster allowed us to queue up render jobs for the entire short before the shots were completely finished so that we could send jobs to the farm the moment they were complete. Muster was great, but could still use some polish. We would have loved a “global queue” feature so we did not have to remember which order jobs were submitted. Also, jobs would appear to complete successfully, when in reality they failed. Every shot required a visual check in compositing or editing to determine if it needed re-rendered. In short, Muster did what it needed to do, and we could not have completed the short without its aid in queueing and managing our render jobs.
After Effects
After Effects saved a lot of render time because it allowed Mike to fix and enhancing a lot of shots in post rather than having to solve minor issues in Maya. Also, After Effects render jobs rendered very quickly, allowing quicker iterations on some of the more complex shot fixes necessary at the end of the short.
Final Renders
Final renders were output as Portable Network Graphics (PNG). This is a loss-less file format that preserves alpha channels. Because the images are generated individually, a render failure does not mean the loss of an entire shot, as would be the risk with video clips.
Premiere
Premiere proved painfully slow when editing our 1998×1080 footage. Because we were unable to preview edits on our project room machines, we had to share time in the editing suite with Dave Kent. This created a few scheduling conflicts and slowed us down as we could no longer pull easily from our networked drives, but the edit suite’s 16 gb of ram (I’m sure that will be laughable some day) provided the speed we needed to edit HD video in real time.
Post Mortem
What we did right
Automation!
-Having render farm submission and certain other tasks (geo-caching) as scripted tools really saved a ton of time. Without them it’s questionable whether we would’ve made our deadline.
Don’t open the scene if you don’t have to
In general the less time you spend opening and checking a scene file will save a lot of time. Try and do as much as you can to a scene via command-line, scripts, or via a spreadsheet. That way you can reduce a lot of tedium at the end of production, when trying to fix shots. We did this rather well for the most part by using a lot of automation and command-line tools.
The Face Machine
Lifesaver for character rigging. Allowed easy control and complex poses to be achieved really fast. Integrating it into our rig scripts was a mere two lines of code and around two hours of set up time, which never had to be repeated. Instant face rig, which would’ve taken a month to complete even a beta version otherwise.
Data Management
Worked really well. Being able to tell which shots were in which stage of the pipeline reduced the amount of shouting around the project room and gave us a clear understanding of what our progress was.
Scripted Rigging
Reduced the amount of time to make rig changes from days/weeks to a few hours at the most once the base rig was complete. Also allowed us to try different solutions to Lenora’s skirt quickly. We were also capable of sharing the same basic rig structure across both characters, without having to re-rig by hand.
Renderfarm
After we added our machines in the project room to the farm we could render everything in the short quickly. We didn’t have to worry about renders failing as much as they did after we split the backgrounds from the characters into passes. Command-Line render setup
Using the command line in Muster allowed us to set up shot resolution and tell Muster which layers to render. This reduced a lot of human error in scene files and fixes could be done quickly, without having to open the scene.
Render Layers
Allowed us to control the renders more in post than in Maya. We could also render the characters without worry about the backgrounds when animation or lighting changes occured. The background passes would stay the same because they were finished, and only the character would be re-rendered.
No cloth, No fur, no crazy VFX
Saved us the time of having more than 4-6 passes per shot,and eliminated a lengthy step from the pipeline. In effect this would have probably doubled the time it took to get a shot to lighting and rendering.
No visual spoken dialogue
Patrick could concentrate on performance rather than waste time lip-syncing a ton of dialogue.
Consistent file structure and naming
Allowed almost anything to be integrated into a script if needed. Since everything followed the conventions writing a script to modify a scene was a matter of concentrating on logic, rather than catching edge cases.
Look-Dev in Maya up front
we got early tests done to see if we could achieve our look in Maya, rather than hoping things would come together later. It wasn’t a final look but it definitely gave us a lot of confidence that we’d keep the look consistent
More complex effects in post
After Effects was able to do more than originally thought in post. Allowing less complex lighting and contour shading in Maya to achieve the Cartoon Modern look.
What we did wrong
Asset tracking
There may be a better more seamless way to integrate our shot progress and which assets were necessary for it, rather than relying on file and directory structure. Perhaps using a show tracking system like Shotgun would help, but it would take some getting used to by the entire team.
Don’t use nested references, or references from other sequences
In some cases a bunch (over 100) cameras were included with some of the clutter assets in sequences 5 and 9, causing huge slowdowns in the scene. Render globals took forever to load and removing the cameras by force would cause scene instability. Ultimately we had to just deal with the lag, but only using top level references could’ve saved us a lot of time. Also, there were times when assets weren’t copied into the sequence they were being used in. They were left in other sequences which the render farm submission script did not process when submitting the shot. It only looked at the sequence of the shot being submitted and the top-level asset tree. Either all assets should have been copied (bruteforce) or assets should have been copied initially for use in that specific sequence.
Have two people for compositing and editing
Crazy amount of work for one person at the end the of the whole process.
Epilogue
Rarely do you get to do exactly what you want with exactly who you want exactly when and how you want to do it. We want to thank the faculty and staff of the ETC who trusted us with their resources, especially our advisers Mk and Ralph, who were our biggest fans even when everything was broken and we seemed lost at sea.
We all feel extremely grateful to have had the opportunity to be a part of StoveGhost Studio and create Collectible. We know it will serve as a calling card for our team as artists and storytellers, but we hope it will also serve as a memorable and thought provoking experience for audiences. Our greatest success will be to see our work living beyond our time together at the ETC.