Manifesto
Radiance fields capture our most cherished experiences. They revive old memories, visualize engineering projects, and augment reality. This isn't a Holodeck fantasy. Industries are already experiencing a renaissance in 3D capture technology.
Yet, cutting-edge advancements remain trapped in GitHub repositories, buried in arXiv papers, or gated by inaccessible compute resources. With millions of new developers entering the workforce annually, accelerated by GenAI, and only a few thousand qualified 3D computer vision engineers graduating, this gap must be addressed.
Unflatten's centralized API for spatial computing tasks frees developers to focus on their craft, not on researching, developing, or hosting complex algorithms.
Our Principles
Make SOTA accessible. By open-sourcing the algorithms behind our API, we ensure they stay current.
Reduce costs through scale. Large-scale scenes are often impractical. Training on large clusters can change that.
Contribute to research. With the data and compute we collect, we aim to lead future breakthroughs.
Our Rejection of the Status Quo
Radiance fields shouldn't require advanced computer vision expertise.
Radiance fields shouldn't be time-consuming or costly.
We reject closed, proprietary software practices.
Our Vision
We envision a future where everyone can create their own personal holodecks. This requires solving inefficiencies in spatial computing. Like LLM APIs such as ChatGPT or Claude, Splatrograph provides an API for generating 3D models without specialized knowledge. By making 3D model creation as simple as text generation, we empower developers to build applications in visualization, simulation, and augmented reality—free from the traditional complexities of 3D reconstruction.
The Path Forward
First 6 months: Foundation
Launch a beta API with core radiance field generation capabilities, supporting small to medium scenes. Open-source initial algorithms and gather developer feedback. Partner with academic institutions to seed research contributions.
Generation time: 2-10 minutes Cost:$1-5 USD
Year 1: Scale
Expand API to handle large-scale scenes, leveraging cluster training to reduce costs. Introduce integrations with major 3D platforms. Publish research advancements driven by anonymized usage data, reinforcing our commitment to the community.
Generation time: 1-5 minutes Cost: $0.10-0.75 USD
Year 2: Ubiquity
Achieve seamless, real-time 3D model generation for AR and VR applications. Enable developers to create custom holodeck-like experiences with zero effort. Establish Splatrograph as the go to API for spatial computing, with a thriving open-source ecosystem.
Generation time: Instant, Cost: Too cheap to meter