23 Daily — According to techcrunch.com, startup Physicl announced on November 6, 2026, that it has secured

0 million in seed funding led by Inovia Capital to tackle the challenge of training robots to adapt from simulated environments to the real world. Founded by CEO Alex de Vigan, whose previous company nfinite created 3D product models for e-commerce brands, Physicl repurposes its prior technology for high-fidelity 3D rendering to improve physical AI models training efficacy.

Physicl bridges the sim-to-real gap with high-fidelity 3D simulations

Physicl aims to solve the difficult problem of aligning simulations with real-world physics in robot training—a critical hurdle in physical AI advancement. Traditional data collection through videos or remote operation is expensive and inefficient. Physicl’s technology transforms unstructured and incomplete real-world information into geometrically, materially, and physically accurate 3D models to generate synthetic training data. Its product Studio offers 10,000 3D environments and 100,000 object assets, enabling users to create scenes targeting complex scenarios—for example, training a robot to identify a cup on a cluttered countertop or under shadow conditions.

Partnerships and dynamic synthetic data generation tools

Physicl’s clients include world model builder SpAItial, which uses Physicl’s high-quality synthetic data to improve the realism of its 2D-to-3D simulation product Echo-3. Beyond being a data supplier, Physicl positions itself as a tooling platform for continuous synthetic data generation. Alex de Vigan envisions customers feeding trillions of images or millions of hours of lidar data into their system, which converts them into detailed 3D representations capable of generating billions of training data permutations on the fly. SpAItial CEO Matthias Niessner, also a robotics professor, praised Physicl’s ability to auto-generate needed data dynamically during training, calling it “super exciting.”

Editor's take: Physicl's technology advances real-world robot applicability and entrepreneurial opportunity

Physicl presents a compelling solution to the costly and time-intensive data collection challenge by rapidly converting multifaceted real-world inputs into high-fidelity synthetic datasets, thereby narrowing the sim-to-real gap in robotic AI training. For entrepreneurs, this underscores promising avenues in building platforms that deliver dynamically generated training data, especially for complex edge cases. Validating Physicl’s impact involves using its Studio tool to produce synthetic data for specific training challenges and assessing improvements in model adaptability, accuracy, and iteration speed. This approach can accelerate the deployment of robots capable of operating effectively in real-world environments, unlocking broad commercial potential.