Physicl, a startup converting e-commerce 3D rendering technology into robot training simulations, has raised $10 million in seed funding led by Inovia Capital.
The company's founder, Alex de Vigan, previously built nfinite, which generated product 3D models and renderings for e-commerce brands. Generative AI saturated that market; de Vigan spotted an opening in a harder problem: robots learning to manipulate physical objects in digital environments that accurately reflect real-world physics.
The core challenge: robot developers typically train on video data or collect expensive real-world examples. Simulation offers a cheaper path to scale, but only if the virtual world resembles physical reality closely enough that skills transfer—what the industry calls the sim-to-real gap.
Physicl's tool, Studio, packages nfinite's geometry-and-physics pipeline into a product for roboticists. It includes 10,000 3D environments and 100,000 assets. Users build scenes to generate synthetic training data for edge cases: a robot identifying a cup on a cluttered countertop, partially shadowed, half-hidden behind other objects.
The startup competes in an emerging segment selling data and simulation infrastructure to robot companies and foundation-model builders. De Vigan acknowledged the market remains nascent and fragmented, with players collecting, labeling, or generating training data across different modalities.
Physicl is focusing on post-training refinement for difficult cases. Early customer SpAItial, a world-model builder, used Physicl's tools to improve fidelity in its Echo-3 product, a 3D simulator that generates scenes from 2D images.
