Dimension scores are derived from public data and fields; weighted into the composite. Reference only.
Splatica is a London-based Physical AI startup co-founded by Andrey Shelomentsev. Its core goal is to build a “data layer for the physical world”: turning a single 360° video capture into a simulation-grade 3D digital twin with physical accuracy, designed for robotics training. The source text indicates that its output can be exported to NVIDIA Isaac Sim and Omniverse, and that it has partnerships or ecosystem affiliations with Insta360, NVIDIA Inception, and others.
Based on publicly available information, Splatica’s key capability is converting 360° video into 3D digital twins. Related media and demos mention 3D Gaussian Splatting, navigable 3D scenes, and the ability to generate walkable scenes from a 3–5 minute 360° capture. It is not a general-purpose text-to-image or text-to-3D tool; instead, it is more focused on data infrastructure for robotics, embodied AI, and simulation training. Typical use cases include reconstructing factories, real-world spaces, drone or panoramic-camera capture scenes, as well as 3D documentation projects for heritage preservation.
The collected source text does not provide pricing, free quotas, trial access, or payment methods, so it is not possible to assess cost-effectiveness or procurement barriers. In terms of integrations, support for NVIDIA Isaac Sim and Omniverse is explicitly mentioned, which is important for robotics simulation teams. Its official Insta360 partner status also suggests that the capture workflow may be well suited to consumer-grade 360° cameras. However, information about APIs, SDKs, enterprise delivery, private deployment, and batch-processing capabilities has not been disclosed. There is also no mention of a Chinese interface, Chinese documentation, or localized support.
Its main strength is a highly focused positioning: it directly addresses real-world scene data needed for robotics training and connects with the NVIDIA simulation ecosystem. Generating a scene from a single 360° capture could, in theory, reduce the cost of data collection and 3D modeling. The limitations are also clear: public materials do not provide key evaluation details such as reconstruction accuracy, scale error, physical-property generation, collider quality, or handling of dynamic objects. The claim of being “physics-accurate” still lacks verifiable metrics. Data privacy is also unclear, including how uploaded materials are stored and whether they are used for training.
Splatica is most relevant for robotics companies, embodied AI labs, industrial simulation teams, digital twin teams, and spatial computing teams—especially users already working with Isaac Sim/Omniverse. For general 3D creators, the available information and access paths may not yet feel productized enough. Access from China is unknown. If the service depends on overseas cloud infrastructure, the NVIDIA ecosystem, or international payments, real-world usage may involve uncertainties around network access, payment, and compliance. Potential comparisons include Luma AI, Polycam, RealityCapture, and Gaussian Splatting-related tools, though these alternatives may not target robotics simulation in the same way.
⚠ This review is compiled from public sources and does not constitute a purchase recommendation. Verify all facts on the vendor's official site. Verify on shelomentsev.com official site.
shelomentsev.com is an Unknown 3D & Assets provider. TG4G tracks its product information, an overall rating of 6.0/10, and a China-accessibility score of Workable. Click "Visit Official Site" to reach shelomentsev.com directly.