Get a free discovery call with Appsilon's expert.
Damian Rodziewicz
Head of Sales
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
Redefining Animation: Rigplay's Journey from Manual to Magical with Generative AI
Experience the future of animation with Rigplay's Generative AI—a game-changer that slashes costs, fuels creativity, and delivers awe-inspiring results.
About Rigplay
Bones Studio, a global virtual production studio renowned for its expertise in full-performance motion capture and character animations for AAA games, collaborated with Appsilon to reshape the animation production landscape.
The animation industry faces cost-related challenges, time-consuming manual processes, and limited creative experimentation opportunities. Rigplay, a generative AI for human animation, addressed these challenges by combining Bones Studio’s industry-leading capabilities with Appslion’s machine learning expertise. In the course of three years, Rigplay has revolutionized animation production by automating processes, resulting in up to 80% cost savings, reduced labor costs, and accelerated creative experimentation while introducing a comprehensive dataset 10 times larger than existing libraries.
Innovating Animation Production through Advanced Machine Learning Techniques
This project leverages generative machine learning techniques inspired by computer vision applications to create human motion animations. Our approach involves animation generation and allows us to control animation styles, such as character expressions (tired, happy) and movement characteristics (masculine or feminine). Additionally, we can generate animations based on textual prompts by combining advanced language models with motion-specific generators.
This is a long-term research and development endeavor, currently spanning three years, aimed at achieving a state-of-the-art level of animation generation. The project is supported by a grant from the National Centre for Research and Development (NCBR).
We have been involved in the project since its inception, assisting in writing the grant proposal, securing the project, and now leading the machine learning component.
Impact & ROI
Appsilon played a pivotal role in the Rigplay project by:
Cost Savings: Drastically reduced costs and time needed to build animation packages and acquire motion capture equipment, saving 5,000 to 100,000 USD.
Labor Cost Reduction: Streamline the animation process, reducing the reliance on specialized animators and saving labor costs.
Customized Animations: Create customized, stylized animations for all characters, not just the main ones, opening new possibilities in virtual worlds.
Accelerated Creative Process: Shift from months to days in animation development, enabling quicker ideation and experimentation, and fostering a more dynamic creative process.
Commercialized Dataset: We co-created a pioneering dataset with over 300,000 animations—tenfold larger than existing libraries—optimized for machine learning, featuring a diverse range of styles, now available as a commercial asset.
Addressing Industry Challenges in Animation Production
In animation production, numerous challenges hinder efficiency, creativity, and cost-effectiveness. Rigplay was initiated to address these critical industry hurdles. The key challenges faced by the animation industry include:
Cost-Effective Animation Modification: Our system minimizes the cost and effort required to modify animations, particularly when altering movement styles.
Affordable Animation Creation: Traditional motion capture methods are expensive, making creating new animation types costly. Our approach, generating animations from textual prompts, drastically reduces expenses while accelerating production. This innovation also fosters creative experimentation.
Speeding Up Animation Preparation: Preparing a set of animations for a 3D character traditionally takes several months. However, our models enable this process to be completed in just days. This remarkable reduction in time translates to increased efficiency and productivity.
The core project activities include:
Curating a state-of-the-art animation database through cutting-edge motion capture technology, resulting in a dataset of over 300,000 animations in collaboration with Bones Studio, our project leader and client. This dataset exceeded existing standards in both size and quality.
Settling on data representation using quaternions, a non-standard data type for ML, but common among animators.
Developing and training machine learning models capable of various animation generation tasks, including style transfer, motion-in-betweening, and text-to-animation.
Delivering the ML models through REST APIs, making them accessible for a prototype application.
Collaboratively creating a web application that allows users to utilize and visualize the outcomes produced by our models.
In collaboration with academic experts, we have designed and implemented evaluation procedures to assess and compare the performance of our models.
The project has achieved significant success and is poised for further development. From a technical standpoint, we employed various machine learning concepts and meta-tools, including:
- Deep Learning
- Auto-encodersVQ-VAE (Vector Quantized Variational Autoencoders)
- Generative Adversarial Networks (GANs)
- Style Transfer
- Transfer Learning
- Unsupervised Learning
- Fine-tuning
- Serialization
- Preprocessing Parallelization
Furthermore, we implemented best practices such as experiment tracking, data versioning, and a continuous integration/continuous delivery (CI/CD) pipeline that includes embedded inference and training tests to ensure the reliability and quality of our work.
“The demand for 3D character animations has been on the rise across various industries such as gaming, entertainment and advertising. And yet, they are very expensive.
Creating them with Generative AI models lowers the cost so much that new possibilities open up, for example, each character in a game can finally move in a unique, stylised way. This is a change which, seen once, will render all previous animations of background characters look artificial.”
Creating them with Generative AI models lowers the cost so much that new possibilities open up, for example, each character in a game can finally move in a unique, stylised way. This is a change which, seen once, will render all previous animations of background characters look artificial.”
Check Data Science Summit Presentation
See Presentation
Take Your Business Further with Custom
Data Solutions
Unlock the full potential of your enterprise with our data services, tailored to the unique needs of Fortune 500 companies. Elevate your strategy—connect with us today!