When the pure-hearted are lonely, they create a world of their own.

@Zhiyan Chen

In 2024, a list of young entrepreneurs made the rounds. Alongside 33-year-old Xingxing Wang, 30-year-old Zhilin Yang, 34-year-old Junjie Yan, and 30-year-old Zhihui Peng, there was a name: Tiange Ling. He was the youngest of them all — just 24 that year.

Then for a while, I stopped hearing updates about him. The venture capital world cycled through a few more hype trains, and I had no idea how his startup was doing.

Until recently, when he reappeared before me.

Five years into entrepreneurship, Tiange Ling has built a murder mystery game app that made a million RMB per month at its peak, created a PC-based AI voice changer, and then pivoted from software into hardware — AI earphones, optimized voice models, and Voice Agents.

We first noticed him not only because he started extremely young, had backing from well-known institutions, and achieved decent results, but also because of another identity: second-generation entrepreneur.

Tiange's father, Hai Ling, was among China's first internet entrepreneurs and former president of Shanda Games. In his youth, Hai was discovered by Tianqiao Chen, moving from Changsha to Shanghai to join Shanda and eventually rising to president — living through the most frenetic years of China's online gaming industry. After leaving Shanda, Hai founded Butterfly Interactive, a gaming company later sold to Baidu. Years later, financially independent, Hai became his son's angel investor.

Hai Ling and Tiange Ling

Yet this isn't the typical smooth-sailing "second-gen entrepreneur" story of a post-00s founder.

Hai Ling introduced Tiange to the world of gaming, gave him his first company, and also became the first person he had to rebel against. In recent years, Hai has been repeatedly admitted to the ICU, and the order originally established under his father's influence began to loosen. Tiange had to bear the fate of a company on his own.

Almost as if by fate, he found himself walking the path his father had left unfinished.

This edition of elselier is Tiange's story.

Whose Startup?

The initial venture began because of the pandemic.

While in his junior year at Carnegie Mellon University, Tiange returned to China for remote classes. The coursework was manageable, so in his free time he worked on side projects with classmates and constantly pitched ideas to his father.

For example: an "anti-Dianping" where people gathered to complain and leave bad reviews; a bill-splitting tool for international students, and so on.

Nearly every one got shot down by Hai Ling. The only one approved was online murder mystery games.

Tiange was already a murder mystery enthusiast. With offline venues closed, he and friends played on Tencent Meeting — "the experience was surprisingly good." At the time, every online murder mystery group chat was full, QQ groups numbering in the thousands. Yet there was no dedicated product for this.

So Tiange founded Halo — an app billed as "the first online murder mystery platform with professional voice actors as game masters." The team recruited大量声优, replicating the offline experience one-to-one online. At its peak, monthly revenue approached three million RMB.

But this venture never seemed fully his from the start.

Tiange had product sense but almost no grasp of corporate governance, business models, or long-term strategy. Hai Ling was not just his angel investor — he also assembled the early team, arranging two of his own longtime partners as his son's co-founders. "In reality, both of them listened to my dad."

Father behind the scenes, son in the spotlight. Looking back, Tiange says it felt more like his father was using him to launch a second act.

Though both wore plaid shirts, Hai Ling was the true captain at first

Unsurprisingly, cracks appeared.

At one meeting, Tiange discovered a partner knew nothing about murder mystery games. He told the person: "You should play a couple games first before working on this."

The response: "You should show me more respect."

Tiange attributed the conflict to different management styles across generations.

In his father's era of entrepreneurship, founders held absolute authority. "(Father) figured things out, then others executed." In that gaming industry, founders secured IPs, copyrights, and distribution channels, judged whether a product would work, then broke targets down for the team to execute. In that system, assertiveness meant efficiency.

But Tiange faced a different problem: What should an online murder mystery game be? How would young people use it? The answers could only emerge from users. He needed partners in the field, not waiting for his orders.

This team tension quickly escalated into father-son conflict. Clearly, Hai Ling saw the capital, experience, and "trusted lieutenants" as support for his son. Tiange, after one argument, said in anger: "If you didn't support me, I could do better."

Before the two generations could settle who was right, the pandemic abruptly ended.

Dead End, Survival, and Turning Point

Players returned offline. Halo's monthly revenue dropped from nearly three million to one million, yet the company had already staffed up like a traditional internet firm — product, operations, marketing, and business development, supporting thirty to forty people.

What truly panicked Tiange wasn't the numbers. It was finally understanding one thing: murder mystery games were good demand, but not a good business.

Users paid for both the scripts and the voice actors, with ARPU even higher than offline — and that was precisely the trap. Of every 1 million earned, 500,000 went to voice actors and 200,000 to script copyright holders.

A two-sided market's marginal costs should decrease with scale, but murder mystery games were a vertical category without enough scale to truly amortize costs.

Strong product-market fit, great user experience, but a fundamentally broken business model.

"Voice" was the core keyword of Tiange's first product, Halo

Worse, Hai Ling fell ill.

It was a drawn-out process. Sometimes he was lucid, able to understand what was happening at the company; sometimes he couldn't. The person who always had judgments and resources to offer gradually withdrew from company decisions.

With the company account nearly empty, Tiange paid all employees' salaries for one to two months out of his own pocket.

Almost in desperation, a lifeline arrived unexpectedly.

During that period, Tiange split the team: thirty people kept the murder mystery game running, while ten worked on a new project — an AI voice changer called "Dabing AI," with the overseas version named Dubbing AI.

The idea didn't come from nowhere. Running the online murder mystery game, Tiange had discovered that good voices commanded extreme premiums. Users didn't just pay for scripts — they paid for the emotion and immersion created by voice actors. When voice matched character, the experience transformed noticeably.

If voice actors could have this power, could ordinary people too?

The ten-person team quickly built a PC software, launching it with extreme crudeness. No marketing, no hype, no paid acquisition — "just put up a website," bought a few search keywords for "voice changer."

Then, a top-tier streamer for Naraka: Bladepoint happened to use it.

The male streamer transformed his voice into a "little sister," pretending to be terrible at the game before suddenly revealing his actual skill. The contrast between voice and gameplay created excellent entertainment value. The chat kept asking: What software is this?

The streamer flashed the product on camera.

That month, two to three hundred thousand users flooded in. More streamers started using it, and the product quickly grew to millions of users.

In the difficult times, Tiange had wanted to lay people off but kept holding back.

"If you restructure when momentum is down — you're already struggling, plus you're cutting people — the company is finished." He kept "hanging on," "hanging on" until the new product exploded and new investment hit the account, then made transfers and optimizations.

Then Tiange told those who remained: "We're going to be okay."

He says from that moment, he truly became a founder.

From Voice to Expression

Soon, "Dabing AI" accumulated over five million users domestically and internationally. Overseas, most were players in Minecraft, Roblox, and Fortnite — games that already had role-playing scenarios; domestically, it was more competitive gamers and streamers. "For the socially anxious, the introverts — give them a voice filter, a virtual voice, and they can speak more easily."

Traditional voice changers mostly relied on adjusting pitch and reverb, only producing unnatural sounds like "Minions" or "Optimus Prime"; another path of "speech-to-text then synthesized speech" tended to lose emotion and add latency.

"Dabing AI" instead used a Transformer architecture for direct voice-to-voice, changing voice in real-time while preserving tone, pauses, even laughter — and reduced early CPU usage from nearly 30% to single digits.

But the PC version soon hit a ceiling.

Voice changing on computers worked because it hijacked the system microphone — something impossible on phones. Yet mobile gamers far outnumbered PC users.

"It wasn't that I wanted to do hardware — I had to," Tiange says.

The question was: how?

In 2024, Tiange set a one million RMB budget for the hardware project: spend it and if it still doesn't work, drop it.

The first generation was a "little tail" dongle plugged into phones. The team deployed models in the box, with users connecting a separate Bluetooth earphone.

This near-"semi-finished" product still sold over ten thousand units. On one hand, it proved demand existed; on the other, it exposed nearly every problem: users had to disconnect their existing Bluetooth, insert the device, download an app, then connect phone, box, and earphone.

More troublesome, the hardware was fully outsourced.

Firmware issues required tickets to suppliers; when sales were good, production couldn't scale immediately; when sales were poor, it couldn't stop. Tiange's product rhythm was measured in weeks; hardware manufacturers had their own schedules.

After the first million was spent, the product was "okay." Tiange poured in more millions.

He hired supply chain and driver engineers, built firmware in-house, sourced components, coordinated with factories, and secured two production lines. The product was reimagined into its most intuitive form: a wired earphone.

Plug into phone, open app, change voice.

To avoid stuffing expensive AI chips into the earphones, the team designed a proprietary protocol: one end connected to the app, with phones handling model computation; the other end used standard protocol to send processed audio back to games, calls, or streams. The team adapted this for nearly five years' worth of phone models.

By end of July this year, these earphones have cumulatively sold over one hundred thousand units. Today, the list of investors backing him is growing — from Source Code Capital, MiraclePlus, to PeakXV Partners (formerly Sequoia India and Southeast Asia), Oasis Capital, and IDG.

But Tiange clearly doesn't want to build a gaming earphone company.

From murder mystery games to voice changers, what he wanted to solve was "what voice to speak with"; after developing accent conversion and emotion enhancement, it became "how to speak better."

What's the next step?

Tiange and his team discovered that even with perfect voice and standard accent, a person still might not know what to say.

"With only a perfect mouth, if the brain can't keep up, it still doesn't work."

After identifying this problem, they first tried having AI prompt users during conversations, telling them what to say next.

This feature quickly failed. No one wanted to chat while reading AI instructions; even if they saw them, they might not follow.

So Tiange inverted the relationship — instead of AI teaching people, let people command AI. This new generation product was called "AI Mouthpiece."

Now, users wearing Dabing AI earphones simply state a goal — help me argue with my opponent, help me liven up the meeting atmosphere, help me strike up conversation with a stranger — and the Agent enters the conversation directly. Users can listen from the sidelines, correct anytime, or take over mid-way.

Among the earliest test users, some chatted with the Agent for a long time first to let it understand them, then said "say whatever you want"; some asked it to learn their speaking style; others mixed different personas, making the language style like Yonghao Luo's with emotional expression like Lin Chi-ling's.

Users originally thought they were training a tool, but eventually turned creating this personality itself into a game.

Tiange told me that internally, the team calls user-stated goals "intention," and the Agent's actual exchanges with others "conversation."

When one conversation goes poorly, users revise their request; the Agent speaks again, users correct again. In this process, what's truly valuable isn't any single piece of audio, but why a person wanted to say something, and whether it ultimately achieved its purpose.

This is also why Tiange believes voice remains undervalued in this wave of AI.

Speech-to-text ultimately returns to text — the battleground major companies know well (which is why he doesn't directly do simultaneous translation). Real-time voice-to-voice is different: it must preserve emotion, tone, and identity; be low-latency enough; and understand different contexts like gaming, calls, and meetings.

In Tiange's view, voice isn't an accessory to text but an interface for AI to enter real relationships.

He believes this product "has a chance to grow into an Agent Layer version of WeChat."

"Directly building another WeChat, you'd definitely lose. But I think an AI-native WeChat solves a different need: letting users make expression their own, based on scenarios."

In his vision, future humans may no longer personally handle every conversation, but instead hand goals to Agents, which find appropriate applications, use your voice, follow your habits, and complete conversations on your behalf.

Online, it might hide in a pair of earphones or an app; offline, it might enter robots, becoming a person's physical avatar.

In this sense, hardware is just a bridge, voice just an entrance. What Tiange truly wants to build is a mouth that understands its owner's intentions and completes the best possible expression.

Father, Childhood, at Shanda

Told to this point, there's a curious loop.

Years ago, after Hai Ling left Shanda, his earliest entrepreneurial attempt was also voice.

Around 2013, Hai built an app called "Taiding" (Arguing). Each day the product presented different topics, and users debated via voice: whether a match was fair, how to evaluate a public event — any issue could spark argument.

What "Taiding" tried to capture was the same human desire to express oneself.

But it was too early. Hai Ling ultimately left voice behind, returning to the more familiar gaming domain. Over a decade later, Tiange started from gaming and walked back to voice.

Two generations entered the same river from opposite directions. Fortunately, Tiange caught the moment when models, compute, real-time communication, and on-device hardware all matured simultaneously.

This story can be traced back even further.

In 2003, Hai Ling was still an executive at a Changsha software company. Shanda was at its zenith, and Tianqiao Chen recruited Hai from Changsha to Shanghai. In the years that followed, Hai became Shanda's president, helping Chen port the PC game Legend to mobile, securing IPs, copyrights, and distribution channels.

But Tiange only pieced together these stories later. What he truly remembers is that Shanghai office at Shanda.

Under his father's influence, Tiange encountered video games at age three or four. Starting at six, he often appeared at Shanda's office. As a senior executive's son, "no one watched over him," so he wandered between different project teams, playing one game after another. "Every game had unlimited resources."

The Ling family's travel style was different too. In every city they visited, father and son would either go to local internet cafes to see what people were playing, or visit GameStop to see which games were selling well.

Hai Ling certainly enjoyed playing games, but he was always observing: why users liked them, why they stayed, why they were willing to pay.

Tiange retains this habit to this day.

There's also an elementary school night that Tiange has remembered for many years.

That day, he and his father were playing Dynasty Warriors at home. The TV faced the bed, both sitting at its edge, clearing stage after stage. Late into the night, one stage simply couldn't be beaten. Tiange, exhausted, told his father: "Dad, I need to sleep." Hai Ling wouldn't let him go.

"What about me if you sleep? We're teammates — only by fighting together can we get through this."

The elementary schooler was prematurely pulled into a "comrade-in-arms" relationship by his father. Only years later did he realize his father wasn't talking about just one game.

In 2010, Tiange accompanied his father on visits to America's best internet companies — a 10-year-old's view of his father at work

Later, Tiange did start a company. But first he had to leave his father's command, replace the teammates his father had arranged, and find his own product and approach.

Hai Ling once told Tiange that a company truly ends not when the account is empty, but when the founder gives up. As long as the person remains, there's always a way. Later, when the company was most difficult, Tiange began repeatedly telling his team this: "There's always a way."

Hai Ling's health deteriorated through this process. Tiange says if his father were lucid now, he would certainly be proud of him. Then he added: "But if he had always been lucid, I might not necessarily be who I am today."

Many years ago that night, it was the father who wouldn't let Tiange sleep, the two of them teammates who had to clear that stage together. Today, Tiange has his own game, his own team, his own voice.

This time, it's his turn to try to wake his father. "Hurry up, look at what I've made."

Cover image: Winslow Homer, Breezing Up (A Fair Wind), 1873–1876, National Gallery of Art