Manycore Tech Open-Sources 3D Gaussian Viewer, Ushering in the Era of the 3D Internet

Manycore Tech has made a significant stride in the evolution of online content by announcing the open-source release of Aholo Viewer, a groundbreaking 3D Gaussian viewer. This innovative tool allows users to explore extensive 3D environments directly from their web browsers, making it compatible with smartphones, PCs, and VR devices, all without requiring any client installation. The Aholo Viewer can render city-scale 3D scenes containing over 1 billion Gaussian splats, which is roughly equivalent to the entire area of Hangzhou's West Lake.

Huang Xiaohuang, co-founder and chairman of Manycore Tech, emphasized the limitations of the current internet, which has predominantly been confined to 2D formats such as text, images, and video for over three decades. He noted that physical spaces are often reduced to mere photographs, and products are represented as flat images, compressing rich physical-world information into lower-dimensional forms. By open-sourcing core 3D infrastructure like Aholo Viewer, Manycore Tech aims to hasten the arrival of the 3D internet era.

Technical Innovations Behind Aholo Viewer

The Aholo Viewer utilizes a streamable Level-of-Detail (LoD) architecture, which allows for the seamless transmission of 3D content across various devices, akin to video streaming. This means that users can instantly access and interact with complex 3D scenes directly from their browsers, regardless of the hardware they are using. This capability is crucial as it democratizes access to 3D content, making it available to a broader audience.

Industry experts have pointed out that the number of Gaussian splats a browser can manage, along with its rendering speed, serves as a critical measure of its effectiveness. Manycore Tech's Aholo Viewer stands out in this regard, supporting up to 1 billion Gaussian splats — a tenfold increase compared to Fei-Fei Li's open-source Spark 2.0 engine, which can handle over 100 million splats on consumer-grade devices. This substantial improvement in both rendering speed and memory efficiency positions Aholo Viewer as a leader in the emerging field of high-performance 3D browsers.

As high-performance 3D browsers become more prevalent, they are anticipated to serve as the primary gateway for mainstream users to engage with 3D content. The experience of exploring immersive 3D environments may soon be as effortless as scrolling through short-form videos, allowing users to seamlessly 'step into' any real or virtual world with just a single link.

Transforming 3D Content Usability

For Manycore Tech, the ability to view 3D content is merely the first milestone. Equally critical is ensuring that this content is usable. In conjunction with the open-source release of Aholo Viewer, the company is also providing APIs for its Aholo Spatial Intelligence Platform. This platform includes features such as spatial reconstruction, cloud rendering, and AI-driven 3D model generation, which aim to alleviate the challenges associated with producing, distributing, and deploying 3D content.

These advancements mean that 3D scenes can transition from being passive viewing experiences to editable and programmable digital assets that can be integrated into various business workflows. For instance, tourism operators could quickly reconstruct entire scenic locations, filmmakers could create virtual production environments, and game developers could design interactive experiences within accurately reconstructed real-world spaces.

Why it matters

The implications of the 3D internet extend well beyond the mere transition from 2D to 3D online content. A more profound significance lies in its potential to address one of the fundamental limitations of artificial intelligence: the challenge of forming a comprehensive spatial understanding of the physical world. Historically, the internet has been dominated by 1D and 2D data, which has constrained AI systems to learn primarily from flat representations. This limitation helps explain why current AI models can produce sophisticated written content yet struggle with spatial reasoning, distorted visual outputs, and basic physical tasks.

As core 3D internet infrastructure like Manycore Tech's Aholo Viewer gains traction, the long-standing barriers to the distribution and accessibility of 3D content may be significantly lowered. If the creation and consumption of 3D content become as straightforward as sharing short videos today, an abundance of real-world spatial data could emerge at scale. This data would be invaluable for training embodied AI systems, including robotics and other intelligent agents designed to function effectively in physical environments. The open-source release of Aholo Viewer thus not only represents a technological advancement but also a pivotal step toward a future where AI can better understand and interact with the physical world.