The miHoYo Story: From Dorm Room to Global Empire

In a modest dormitory at Shanghai Jiao Tong University during the spring of 2012, three computer science undergraduates—Cai Haoyu, Liu Wei, and Luo Yuhao—pored over lines of code that would eventually power a global entertainment empire. Their venture, miHoYo, began not with a grandiose business plan but with a simple, fervent belief: that technology could be the ultimate enabler of artistic expression in games. This philosophy, born from late-night discussions fueled by cheap takeout and a shared love for Japanese anime and role-playing games, set them on a path to industrialize the very process of imagination. What follows is the chronicle of how three students transformed a dorm-room project into HoYoverse, a studio whose titles like Genshin Impact and Honkai: Star Rail now define a generation of cross-cultural gaming. Their story is a masterclass in scaling passion through technological innovation, long-term vision, and relentless execution—a blueprint for any young creator dreaming of worldwide impact.

The Dorm Room Genesis: Bootstrapping a Vision (2012–2014)

The origins of miHoYo are rooted in the quintessential tech startup narrative: three friends, a cramped space, and an audacious goal. Cai Haoyu, the de facto technical leader, had been immersed in computer science since childhood, writing his first game at age ten. Liu Wei, the artistic visionary, brought a deep appreciation for anime aesthetics and character design, while Luo Yuhao handled business development and operations, ensuring the fledgling studio stayed afloat. Their initial capital consisted of personal savings and a modest student loan—roughly 100,000 RMB (about $16,000 USD at the time)—which they used to rent a tiny office in Zhangjiang High-Tech Park after graduating, though the earliest days were spent in the dorm, coding through the night.

Their first project, *FlyMe2theMoon* (2012), was a small-scale iOS game inspired by classic arcade titles. It served as a proof-of-concept but failed to gain traction commercially. The team realized that to survive, they needed to target a niche where their passion could translate into a dedicated audience: the Chinese “ACG” (Anime, Comic, Games) community. This led to *Houkai Gakuen* (2013), a 2D side-scrolling shooter based on a webcomic concept, which garnered a modest following. The real turning point came with *Houkai Gakuen 2* (2014), a mobile iteration that refined the core loop with gacha mechanics, vibrant anime-style visuals, and tight combat. Built on the Unity engine—chosen for its accessibility and cross-platform potential—the game’s success was immediate, generating enough revenue to pay the first salaries and move the team out of the dorm. By the end of 2014, miHoYo had grown to about a dozen employees, still tiny by industry standards but now financially self-sustaining. The lesson here was clear: technological agility (rapid prototyping in Unity) combined with deep cultural insight (authentic ACG aesthetics) could create a product that resonated powerfully with a specific audience, providing the foundation for scaling.

The Technical Crucible: Honkai Impact 3rd as a Showcase (2014–2016)

*Honkai Impact 3rd* (2016) was miHoYo’s declaration of technical ambition. The team set out to prove that a mobile device could deliver a 3D action RPG experience comparable to console titles—a daunting challenge given the hardware constraints of 2010s smartphones. The project required a complete overhaul of their technical approach, moving from 2D sprites to fully rigged 3D models, complex animations, and real-time physics.

At the heart of this leap was a deeply customized Unity pipeline. While Unity provided the foundation, miHoYo’s engineers wrote extensive custom code to overcome mobile limitations. One critical area was character rendering. To achieve the distinctive anime aesthetic—clean lines, exaggerated expressions, and dynamic lighting—they developed a proprietary cel-shading pipeline that went beyond standard toon shading. This involved:

• **Outline detection**: Using a combination of depth-buffer inversion and edge-detection shaders to create consistent black outlines around characters, essential for the comic-book look.
• **Specular control**: Traditional physically based rendering (PBR) produces realistic but often muddy highlights; miHoYo implemented custom lighting models where specular spots were small, sharp, and artist-controllable, mimicking hand-drawn sheen.
• **Subsurface scattering approximation**: For skin, a simplified scattering effect simulated the soft glow of light passing through ears and cheeks, achieved via texture-space diffusion without heavy ray tracing.

Animation posed another monumental challenge. *Honkai Impact 3rd* features fast-paced combat with fluid combos, dodges, and ultimate abilities. The team built a state-machine-driven animation system layered on top of Unity’s Mecanim, allowing seamless transitions between hundreds of animation clips based on player input

...input and combat events. This system, which the team internally dubbed "BattleFlow," supported the game’s signature "Ultimate" abilities that could trigger cinematic camera angles and slow-motion effects without breaking immersion. The result was a combat feel that, on a 2016 smartphone, mimicked the responsiveness of a console action title—a feat that earned *Honkai Impact 3rd* critical acclaim as a "tech demo come to life."

Beyond animation, the physics and visual effects systems were equally groundbreaking. The team implemented a rigid body dynamics engine built on top of Unity’s PhysX integration, enabling destructible environments— crumbling pillars, shattering ice—that reacted to character attacks. For the iconic "Burst" mode, where characters unleashed screen-clearing attacks, a custom particle system managed thousands of simultaneous sprites with GPU instancing, a technique then rare in mobile gaming. Audio design, often overlooked, received equal attention: a dynamic music system using FMOD middleware seamlessly transitioned between combat, exploration, and story tracks based on in-game state, heightening emotional beats. These systems weren’t just bolted on; they were architected to work within strict mobile constraints—thermal throttling limits, memory budgets under 500MB, and variable network conditions. The engineering culture emphasized profiling-driven optimization: developers used Unity’s Profiler and custom in-house tools to identify bottlenecks down to the draw call level. For instance, reducing character shader complexity from 70 to 45 instructions per pixel freed up enough GPU headroom to add real-time shadows without dropping frames below 60 FPS on flagship devices.

The game’s live-service model introduced new technical hurdles: server-authoritative combat to prevent cheating, hotfix deployment for balance patches, and content pipeline scalability for regular updates. The networking stack, built on a custom C++ layer over TCP/UDP, synchronized player inputs with server-side validation, a necessity for competitive leaderboards. Early versions suffered from latency spikes during peak hours in China, prompting the team to implement regional server sharding and predictive client-side reconciliation—a technique borrowed from PC FPS games. Content updates, delivered every 6–8 weeks, required a robust asset bundling system that could push new character models, levels, and story chapters as incremental downloads (often under 200MB) without forcing full game reinstalls. This pipeline, while primitive compared to today’s standards, became the bedrock for miHoYo’s future live-service operations.

*Honkai Impact 3rd*’s success—over 10 million downloads by 2017 and a cult following overseas—validated the thesis that mobile hardware could be pushed to deliver console-like experiences through software ingenuity. More importantly, it funded the next leap: an open-