Disney Research and Pixar Animation Studios deployed a machine learning-based denoising system within Pixar's RenderMan rendering software, which dramatically speeds up the production of animated films by removing visual noise from partially rendered images without requiring days of additional rendering time.

The denoiser is a deep learning model trained on millions of rendered image examples from Pixar films, including 'Finding Dory,' that learned to convert noisy, quickly rendered images into clean, production-quality frames. This technology was integrated directly into Pixar's RenderMan pipeline and has been used in productions including 'Cars 3,' 'Coco,' and 'Elemental.' Disney Research also explores machine learning for in-betweening animation frames — generating intermediate frames between two keyframes for non-character animation. The RenderMan denoiser is described by Pixar as enabling the studio to achieve high image quality in a fraction of the rendering time previously required.