Agenci AI · karta decyzji
Lyra 2.0 Nvidia
NVIDIA
Lyra 2.0 (Nvidia) is an open-source AI framework that generates persistent, explorable 3D worlds from a single image or video, enabling real-time exploration via browser-based Gaussian Splatting and export to NVIDIA Isaac Sim for robotics simulation.
Dostęp
Open source
Cena od
unknown
Trial
unknown
Sprawdzone
9.08.2026
W skrócie
Czy to narzędzie pasuje do Twojego zadania?
unknown
Zastosowania
- Creating interactive 3D environments for games and simulations
- Generating persistent 3D scenes for VR/AR applications
- Building synthetic training data for robotics and autonomous systems
- Developing real-time 3D content for marketing and social media
- Research in generative 3D modeling and spatial consistency
Najważniejsze funkcjonalności
- Generates persistent, explorable 3D worlds from a single image
- Supports long-range video synthesis with strong global geometric consistency
- Reconstructs generated video sequences into explicit 3D representations
- Enables real-time rendering of 3D worlds via browser (desktop, mobile, VR)
- Uses Gaussian Splatting for 3D rendering with support for over 100 million points
- Supports export to Gaussian Splatting or meshes for use in NVIDIA Isaac Sim
- Designed to prevent spatial forgetting through per-frame 3D geometry routing
- Trained to correct temporal drifting using self-augmented histories
- Open-source framework based on Three.js for browser-based 3D visualization
Darmowy plan i trial
Darmowy dostęp: unknown
Trial: unknown
Cennik: unknown
Dane i bezpieczeństwo
- Prywatność
- unknown
- Trening na danych
- unknown
- Region danych
- unknown
- RODO / GDPR
- unknown
- Bezpieczeństwo
- unknown
Materiały i dowody
unknown
Materiały producenta
Badania, benchmarki i raporty
Tytuł materiału producenta nie jest niezależnym dowodem skuteczności. Źródła oznaczamy według ich pochodzenia.
Mocne strony
- + Open-source and accessible via Hugging Face
- + Supports real-time exploration of 3D worlds in any browser
- + High scalability and spatial persistence in scene generation
- + Optimized for NVIDIA GPU-accelerated systems (H100, GB200)
- + Enables integration with robotics simulation via NVIDIA Isaac Sim
Na co uważać
- Relies on synthetic data for training, which may limit generalization to out-of-distribution scenarios
- May inaccurately reconstruct videos not present in the synthetic training domain
- Not suitable for public distribution, deployment, or commercial use per license terms