Graphics veteran Tong Xin joins Meshy as chief scientist

3D generation is shaping up as the next high ground for AI innovation, and the latest signal comes from the people building it. According to a QbitAI report on September 17, Tong Xin, one of the most senior figures in computer graphics, has formally joined Meshy — the AI 3D company founded by Hu Yuanming — as chief scientist, responsible for setting its long-term research strategy.

Tong spent 25 years at Microsoft Research Asia, where he led the Internet Graphics group and served as a global research partner. His research spans computer graphics and 3D computer vision, including material capture and modeling, texture synthesis, 3D geometry processing and modeling, light transport analysis and simulation, realistic rendering, and 3D facial animation.

He joined Microsoft Research China in 1999, freshly out of a PhD at Tsinghua University, as one of the lab's first researchers, and eventually rose to global research partner. According to QbitAI, his publications have accumulated more than 21,000 Google Scholar citations, and his work fed into technologies including Xbox game development APIs, Xbox-compatible software, Windows 3D printing drivers, and the Direct3D graphics development toolkit.

Graphics is the discipline of creating, manipulating, and displaying three-dimensional worlds on a computer, which makes it foundational to games, film, and industrial simulation. It is now also the infrastructure for how AI perceives and expresses the physical world: without an understanding of three dimensions, an AI cannot really enter physical space.

Meshy was Hu's first company after his PhD at MIT, built around a simple proposition — turning text and images into 3D models. It has gathered 12 million users, and its customers cover half of the ten most valuable companies in the world by market cap and valuation. In July, it closed a Series B of nearly $400 million at a post-money valuation above 10 billion yuan, setting records for both single-round funding size and valuation in the AI 3D sector.

Hu's stated ambition goes further than model quality. He calls it "AI for Fun," reasoning that once AGI solves most productivity problems, humanity will be left with how to create, express, connect, and find meaning — and that AI-driven joy and meaning-making will be one of the most important problems of the next five years. Meshy is pursuing three layers of fundamental work: reinventing graphics to escape the local optimum of triangle-based rendering pipelines; building a "director system" that defines how a world runs and writes its script in real time; and low-latency, high-fidelity, low-cost 3D generation and rendering infrastructure.

Tong's own research agenda overlaps with that direction. As early as 2016 he framed the goal as graphics content production for everyone and everywhere; in 2024 he posed the question of whether 3D is merely a special case of video generation. Meshy's Meshy-7 release on August 10 pushed geometry alignment forward, restoring facial expression, anatomy, and skin detail in characters while placing mechanical parts accurately in hard-surface models.

As the QbitAI piece was being written, Hu published new team work called Mora — Multimodal Open-world Real-time Architecture — composed of a coding agent that generates world skeletons and runtime code, Meshy's 3D generation that enriches those skeletons and outputs control signals, and a video model that produces final visuals and sound. Hu calls it a technology that goes "beyond world models," though Mora 1 remains at an early framework-validation stage.

For Tong, the moment is familiar: neural rendering, generative AI, and now video models have each challenged the boundaries of classical graphics. Whether this pairing of two generations of Chinese graphics researchers can turn a single sentence into an enterable, interactive world will depend on how 3D generation and video models divide the work — a question worth watching closely.