In the realm of 3D content creation, a silent revolution is unfolding. With the surge of interest in "Black Myth: WuKong" sweeping through the gaming industry, AI-driven 3D generation technology is quietly rising, garnering widespread attention in the industry.
In this global competition for large-scale 3D models, domestic team VAST has showcased China's innovative prowess in this field with their latest creation, Tripo2.0.
Founded last year, VAST is led by CEO Song Yachen, who previously oversaw multiple AI projects at SenseTime, and CTO Liang Ding, an alumnus of Tsinghua University. The company is dedicated to creating a user-friendly platform for 3D content creation and has recently secured hundreds of millions in funding, setting a new record in the field of large-scale 3D models.
The core advantage of Tripo2.0 lies in its powerful generation capabilities and astonishing efficiency. It can generate geometric previews in seconds and quickly complete the creation of textures and materials. Our practical tests have been impressive: whether it's generating half-body anime girl images, full-body cartoon figures, or complex monster designs, Tripo2.0 can deliver high-quality 3D models rapidly.
Particularly noteworthy is Tripo2.0's exceptional performance in converting images into 3D models. We tested it with an ice cream image, and the resulting 3D model was flawless from every 360-degree angle, with stunning color and texture reproduction.
The success of Tripo2.0 is离不开 its underlying technological innovations. It employs the 3D Scaling Law, trained on massive high-quality 3D data, and successfully captures the distribution characteristics of geometry and materials through probabilistic generative modeling methods.
In terms of architectural design, Tripo2.0 cleverly combines the strengths of DiT and U-Net, with the former excelling in capturing global information of 3D structures and the latter in preserving details.
This model not only generates high-fidelity PBR materials but also accurately captures the features of the original images, maintaining complete details from the front, side, and back. More remarkably, it supports negative prompt inputs and even allows users to customize model proportions and poses, demonstrating strong controllability.
The launch of Tripo2.0 marks a significant breakthrough in domestic 3D generation technology, opening up new horizons for content creation. It not only greatly enhances the efficiency of 3D model generation but also provides a convenient creation path for ordinary users without professional skills. This technology is expected to play a crucial role in various fields such as game development, film special effects, and virtual reality, driving innovation and development in the content industry.
Official website: https://www.tripo3d.ai/
