Image Credit: Model and animation are downloaded from Mixamo: Adobe. 2024. Mixamo. https://www.mixamo.com/ Accessed: 16-01-2024
Featured in the SIGGRAPH 2024 Technical Papers, “SMEAR: Stylized Motion Exaggeration With ARt-direction,” by Jean Basset, Pierre Benard, and Pascal Barla brings the dynamic expressiveness of 2D smear frames into 3D animation. This Blender-integrated tool combines beginner-friendly presets with advanced customization, tackling longstanding challenges in 3D workflows. It empowers animators to effortlessly prototype and refine motion, opening new creative possibilities.
SIGGRAPH: Share an overview of “SMEAR: Stylized Motion Exaggeration With ARt-direction.” What inspired this research? Were there specific challenges in traditional animation techniques that sparked this innovation?
Jean Basset (JB), Pierre Benard (PB), and Pascal Barla (PB): This research is part of the MoStyle ANR project. The goal of this project is to investigate how computer tools can help capture and reproduce the typicality of traditional 2D animations. Specifically, recreating the unique look and feel of traditional 2D animation in 3D is a difficult task, due to the computational rigid look and feel of 3D animation. One approach is to mimic, in 3D, motion stylization techniques found in 2D animations to break this rigid feel. In traditional 2D animation, smear frames are specifically aimed at exaggerating motion by drawing elongated objects. With the recent popularity of 3D-animated films with strong stylistic identities, such as the “Spider-Verse” movies, these motion effects have gained increased interest. However, these effects are more difficult to create in 3D than in 2D, since animators need to deform 3D objects. This often results in tedious manual approaches making iterations difficult, hence the interest of a semi-automatic tool.
We grounded our work on observations and discussions with artists which guided the design of our tool. We studied, for example, traditional 2D animations that introduced smear effects, such as “The Dover Boys at Pimento Academy” (Chuck Jones, 1942), and educational resources for animators such as Richard William’s “The Animator’s Survival Kit” book. We also conducted a one-hour interview with a professional animator and had informal discussions with several others. This allowed us to identify real artistic challenges and to define guidelines with artists for the creation of stylization tools.
The result is an art-directable method that allows us to create and interactively customize the main kinds of smear frames: elongated in-betweens, multiple in-betweens, and motion lines, applied to 3D animated objects and characters.
SIGGRAPH: The tool offers preset stylization functions for beginners and more advanced customization for experienced artists. How did you balance these needs to make the tool accessible yet powerful?
JB, PB, and PB: The stylization step of our method is based on a multi-level authoring tool (inspired by Ma et al. 2022). We propose stylization functions with high-level parameters, accessible through simple controls such as sliders, usable by novices and allowing fast prototyping. The stylization functions are implemented in a node-based interface, which is an interaction metaphor that most experienced 3D artists are familiar with. This allows them to customize existing stylization functions or to create entirely new ones.
In addition, we propose a Blender implementation of our method (available directly in the add-on library!). Blender is a popular, free, and open-source software, making it easy to experiment with our method and to integrate it in artists’ workflows. Blender is already used in the industry, especially in small studios. It offers its own powerful node network system to manipulate and generate geometric primitives: Geometry Nodes, in which we developed our method. Geometry Nodes can read information from Blender, such as painted weights as demonstrated in the paper, allowing more advanced use of the tool.
SIGGRAPH: Your paper mentions achieving interactive rates for stylized motion exaggeration. Could you explain how this real-time aspect enhances the animation and refinement process for artists?
JB, PB, and PB: An important point found in our discussions with artists is that stylization tools must be easy to integrate into the animators’ workflows and allow customization. This makes interactive visualization of the stylized results important, since it allows fast prototyping and iterations during the animation process. This is in contrast with simulation-based approaches that can neither be evaluated instantaneously at a given frame (due to temporal dependencies between frames), nor directly controlled/edited by the animator, hence often applied as a post-process and requiring dedicated technical artists.
SIGGRAPH: Are there any additional smear frame effects or stylistic tools you’re considering for future updates? How do you see this work evolving over time?
JB, PB, and PB: There are several aspects on which we plan to work in the short term. First, our tool does not give satisfactory results on complex animations with overlapping trajectories. For example, stylization of animations with important rotation (such as a spin top) or with impacts (such as a ball bouncing off a wall) can create severe (self-)intersections. We are studying how artists stylize such motions, and we aim to propose new stylization approaches to adapt the SMEAR system to such cases. Another direction we are considering is perceptually guided stylization. As highlighted in our discussions with artists, smear frames are often used to guide the observers’ gaze. We aim to study this effect to better understand how artists convey motion to the observers through smear frames and use this to control stylization. Finally, an interesting future direction would be instantaneous motion offsets computation for real-time applications such as games.
Another important evolution of this work will be to obtain feedback directly from professional artists using our implementation to identify avenues for future work. We recently published an open-source implementation of the tool as a Blender add-on. This add-on has been downloaded more than 2,000 times in a month since its publication in early December, and we started contacting studios and artists to initiate direct collaborations. In the long term, we plan to continue offering support on this add-on based on real users’ feedback and to extend it based on new research results.
SIGGRAPH: Your Technical Paper was also featured as a demo in the SIGGRAPH 2024 Emerging Technologies Experience Hall. How did participants interact with your demo? Do you have any key reactions or takeaways to share from the experience?
JB, PB, and PB: During the demo, we presented the Blender implementation of the method. The participants could experiment with the preset stylization functions, and we presented how the tool could be used to create and customize new styles. Participants showed great enthusiasm and were very curious to see how artists would use the tool in the future. The demo was a great opportunity to discuss in depth with SIGGRAPH attendees. We would recommend all authors with working prototypes of their work to participate in the demo sessions!
While the SIGGRAPH 2025 Technical Papers program may be closed, there’s still time to bring your ideas to life! Explore the wide range of SIGGRAPH 2025 programs still accepting submissions and find the perfect home for your work.
Jean Basset is an associate professor in Computer Graphics at Université de Bordeaux and researcher in the Manao team (Inria Bordeaux, CNRS, LaBRI). His work focuses on 3D animation, specifically motion stylization and artistic control, and motion perception.
Pierre Bénard is an associate professor at the University of Bordeaux and a researcher in the Manao project team shared by LaBRI (Bordeaux Computer Science Research Laboratory) and Inria. His research focuses on expressive rendering, which aims to define tools for creating and editing stylized images and animations.
Pascal Barla is a research scientist at Inria Bordeaux in the Manao team. His research focuses on expressive rendering, perception, material appearance and vector graphics.
