5 minute read
AI: Friend or Foe?
By Jose Frederick San Román
Wonder Park
A look ahead at the applications of artificial intelligence in the animation process.
The world of 3D CG animation has grown rapidly over the past few decades with larger production houses and more content generated across the board. In addition to the U.S. major animation studios and a few large independent players, VOD streaming companies are now powerful players in this space as well. Furthermore, the current global pandemic has led to a sharp rise in the consumption of animation content to unprecedented levels, putting greater demand on animation production houses to revamp their production pipelines and deliver yet more content at a global scale. One of the popular buzzwords of our current era is Artificial Intelligence (AI), which will have an important role to play in the animation space. By definition, AI is a broad term that includes many existing and upcoming new tools that will interact between humans and technology (hardware and software). It has only been in the past 10 to 15 years, with the increase in processing speeds and more data available, that AI has really taken off. As such, AI in the animation industry is also going to disrupt how production pipelines work, speeding up the turnaround of projects, shortening production timetables and allowing artists and talent alike to focus more on the creative part of content and less on the production process itself. Contrary to popular belief, AI is not here to take your job. In fact, AI consists of tools that will “democratize animation production” by lowering barriers to entry as production pipelines become a commodity and are easily accessible at a fraction of the cost a decade ago. Both the creative community as well as technical talent should embrace this new wave as an opportunity to really work on generating new and exciting original stories in yet unimaginable visual settings.
Pushing Efficiency
For the time being, however, the actual process of producing animation is still very much manually driven by the artists themselves, which makes it unreasonably expensive and time consuming. The key to introducing AI into an animation pipeline is to drive production efficiency while retaining the high artistic quality produced by the artists, but at a fraction of the time (and cost) spent in the past. Whereas a number of AI tools are in the process of being developed as we speak, three come to my mind as key for an animation pipeline to thrive in this brave new world where there is a higher demand for output and more original content. They are: Robotics Process Automation (RPA), Machine Learning (ML) and Natural Language Processing (NLP). Below is a classic animation pipeline where potentially AI can be introduced.
Mortadelo and Filemon
Robotics Process Automation or RPA is the first of the AI tools that can make a significant impact in numerous departments of an animated production pipeline, as shown in the diagram above. For example, in the “Sets & Props” department, RPA can help in “building” the library of many objects that will be used in the feature (i.e. car, lamp, book, plate, road, tree…), once the production designer has established a “look & feel” for the production. This unit alone, integrating AI RPA tools, can significantly cut the time to process these early stages of production work, saving money and time. RPA can provide 80-90% of the standard movements a character (be it human, animal or living-object like) will make. Movements such as limbs (hands, fingers, legs) or hair (very complex) are generated and stored in a library that will “improve” as more “animation memory” adds to the prior process. The artist will ultimately focus on adding their personal talent to finessing these RPA generated movements, again at a fraction of time and cost in a production. Machine Learning (ML) focuses on automating a process by finding a solution based on multiple recurrent examples. Multiple variations of ML exist (supervised, unsupervised, active learning, transfer learning, etc.). In a classic animation pipeline, perhaps the biggest impact will be when ML is used jointly with pipeline management programs such as Shotgun. ML can help in automating, approving and streamlining the decision making process, allowing artists (and production crew) to spend greater time on the creative direction of the project and its storyline. As the AI tool develops over time with better quality data and recurrent intuitive interaction, the production pipeline improves in efficiency and turnaround times, impacting positively both on the shortened delivery date and reduced production costs derived from savings across the whole process. Natural Language Processing (NLP) could be considered a variant of ML within the AI universe. In an animation pipeline, NLP can make an important impact when combined with ML tools. Assuming we have a closed script, an agreed production design, character specs and sets & props already in place, NLP used together with ML can teach the computer to “read” the script, based on a given context defined previously in my earlier four elements, and then “act” in order to “animate” the given character or object (i.e. arm, legs, helicopter, car, cloud in the sky, etc.).
Close Monitoring Required
The introduction of AI tools in an animation pipeline should be done in small steps, ensuring close monitoring to generate key performance metrics that can substantiate the expected efficiencies and cost savings established at the outset of the process. If successful, artists can focus more on generating ideas and creating new worlds, and less on the more mechanical process of the feature film. This entry of AI in the animation space should boost the number of original content developed and allow for more production players to enter the game with lower barriers of entry. Time will tell, but early tests indicate AI will have a positive impact in the animation process, which will allow for more players to enter and grow the talent pool worldwide. Special thanks to Thomas Malone (Patrick J. McGovern [1959] Professor of Management at MIT Sloan School of Management), Daniela Rus (Andrew [1956] and Erna Viterbi Professor of Electrical Engineering and Computer Science; Director of the Computer Science and Artificial Intelligence Laboratory [CSAIL] at MIT), Brian Charles and Rodwell Mangisi for their support in the field of AI. ◆