We designed our technology to be performer-friendly and scalable, enabling content creation without the need for large motion capture stages.

Currently, we manually build the component catalog for synthetic data generation, but we are developing AI tools to automate this process and intelligently assemble components into systems.

In our projects, we aim to create products that are computationally, robotically, and materially novel, transitioning from traditional assembly to using robots and microbes.

To create characters like Morales, we developed a unique virtual production tool that combines live streaming video and audio with motion capture technology, allowing us to build a storytelling studio.

We created an 'arm farm' where robots learn to grasp objects by trial and error, improving their grasping abilities through pooled sensor data.

We haven't used AI in final production yet, but it's incredible for prototyping and imagining scenes.

By running 10,000 simulations in parallel, we can transfer training to real robots without any fine-tuning, showcasing the power of simulation in robotics.

To train robots effectively, we must simulate environments at 10,000 times faster than real-time and vary parameters like gravity and friction to ensure adaptability.

OpenAI o3 has automated three to four steps of a normally lengthy workflow, generating a report that would typically take hours to complete.