The author also discusses the use of AI in raster and vector frame representation. For raster frame representation, the author mentions a paper titled "Improving the Perceptual Quality of 2D Animation Interpolation" (2022) which uses forward-warping based on bidirectional flow estimation. For vector frame representation, the author refers to a paper titled "Deep Geometrized Cartoon Line Inbetweening" (2023) which formulates inbetweening as a graph matching problem. The author concludes that while AI can improve animation workflows, it requires careful application and tailoring to specific needs.
Key takeaways:
- The author is exploring the use of AI in 2D animation, specifically in the role of an assistant to improve productivity in traditional hand-drawn animation workflows.
- The focus is on "inbetweening" - the automatic generation of intermediate frames between given keyframes. The author reviews two recent papers on this topic and tests a commercial frame interpolation tool.
- The author also discusses the future of these techniques versus other emerging approaches, and invites feedback and collaboration from others interested in this field.
- Two different approaches to AI in animation are discussed: raster frame representation, which treats the input as images, and vector frame representation, which works on a vector representation of the lines. The author tests both approaches on two animation sequences.