Photorealism is achieved by using rendering systems to simulate the real-world flow of light. The more light rays the better. However, simulating light rays is processor-intensive and there never seems to be enough rendering resources to move through a given week’s inventory. We have the option to render at lower quality and use fewer rays, but the resulting noise makes these elements unusable for the final film.
Artificial Intelligence and Deep Learning are technical trends that will have an impact on this issue. Researchers at UC Santa Barbara, Disney and Pixar are creating software that trains a deep-learning model called a convolutional neural network. The system learns how to transform noisy images into renders that resemble those computed with more light rays. This technology will enable animation studios to create more high-quality content – faster and with fewer rendering resources.
Animation studios will also leverage online, cloud-based resources to meet increasing demands for content. ‘Studio-in-abox’ services like Foundry’s Elara will enable storytellers to access software, data storage and rendering resources on-demand from any web-connected device. Independent productions and even solo artists will have access to cutting-edge technology and limitless computing power without having to shoulder the burden of building a pipeline and purchasing expensive hardware infrastructure.
Imagine a world where a group of artists crowd-fund an idea for an animated feature film. They use the money to access cutting- edge software and hardware resources. They build assets, animate, light their scenes and send them off to a distant render farm that is quietly teaching itself how to make images more beautiful, more quickly.