Physicl raises $10M to make digital worlds realistic enough for robot training

De Vigan spent the last decade building nfinite, which helped brands and retailers create 3D models of their products for online display. The arrival of transformer-based AI models led that company to focus more on generative imagery, but de Vigan said he saw a new purpose for the team and the tech stack built to faithfully render 3D environments. "What we were super unique about doing in 3D, AI was going to replace, but something was missing in AI: its ability to understand, interact, and manipulate space and objects," he says.

Researchers developing physical AI models to make robots capable of tasks like handling diverse, small objects have been frustrated by the returns of training them on video, and by the expense of gathering data through remote operations. Simulations in digital worlds are seen as a promising solution, but ensuring those worlds are faithful to real-world physics has proven difficult. De Vigan believes the pipeline his previous company built to "transform unstructured, incomplete, heterogeneous real-world information into a high-fidelity physical representation… with correct geometry, materials, and physics" can now be applied to physical AI.

Physicl is entering a crowded category of startups vying to provide data and tooling to foundation model builders and robot operators. “The space is very young and very blurry sometimes,” de Vigan admits. Companies in the category are collecting, labeling or generating data in different modalities while customers search for the approach that makes robots live up to their promise. Physicl wants to focus on post-training models on difficult edge cases, and Studio lets users create scenes to generate synthetic data for specific problems. For example, Studio could be used to generate training data to help a robot that has trouble discerning a cup on a cluttered countertop or in shadow.

SpAItial, a world model builder, collaborated with Physicl as an early customer to improve the fidelity of its products, the latest of which, Echo-3, was released last week. SpAItial's tool generates 3D simulations from 2D imagery, and its CEO Matthias Niessner, also a professor of robotics at the Technical University of Munich, said the company needed to train on the kind of high-fidelity synthetic data Physicl offers. "The real challenge is, you have to curate the data by hand, and this is a very long cycle," Niessner told TechCrunch. "To us what seems very attractive is you can just do this on the fly […] You're running Claude, and it checks, 'Oh, I need more data of this in my training,' and then it auto generates that data, depending on the training state. That, to me, is super exciting, and this is what Physicl is offering."

Physicl wants to be more than a data provider, positioning itself as the tooling for ongoing synthetic data generation. De Vigan said what his team created 10 years ago was, without knowing it, "basically an agentic workflow to transform unstructured information into high fidelity 3D." He imagines customers bringing trillions of images or millions of hours of LIDAR capture data into the platform to turn into high-quality 3D representations, generating billions of permutations to train their models.