WAIC Postscript: Beyond the Narrative of Large Models, An Unconventional Start for a Company With Hundreds of Millions of Users
The attitude a company holds towards AI today may be a portrayal of its values.
The recently concluded WAIC has demonstrated the current booming momentum of the AI industry. Countless tech practitioners gathered in Shanghai. Even scalpers' tickets were priced sky-high yet still hard to get, and neither the sweltering heat nor thunderstorms could stop people's enthusiasm for pursuing changes amid the AI wave.
The attitude towards AI variables is not only a choice for individuals, but also a key factor that enterprises need to weigh. The attitude with which they pursue changes in the AI wave also determines their posture to face the future.
At this year's WAIC, an AI guitar drew widespread attention. Named "TemPolor", the world's first generative AI guitar allows users to play songs along with the rhythm of light and shadow the moment they pick it up, no music theory foundation required.
Behind this guitar is Timi Tech, an internet technology company that once created products with a user base of 200 million. At present, Timi Tech has realized an integrated full-stack layout covering "model-application-hardware" in the AI music track, making it the first manufacturer in the industry to complete such a layout.
In Timi Tech's AI vision, technology should always serve the essential needs of human beings, rather than create estrangement or replace people. For Timi Tech, which started its business with mobile voice services, it has built an ecosystem centered on voice, connection and interest-based social interaction for Generation Z, among which "the companionship between people" is its most valuable asset.
Based on this, it chooses to develop new businesses in vertical scenarios it excels at. Long Ling, Co-founder of Timi Tech, told 36Kr that Timi Tech currently has corresponding businesses in AI music, AI dubbing and translation, AI 3D and other fields, with the core focus on its strong digital entertainment and cultural scenarios, to create AI products and services that meet the needs of people with diverse interests.
"If our goal is to achieve long-term growth in the AI era, we should see that the value of AI lies in its ability to bring more groundbreaking user experience and business models. I need to evolve our company from a company with AI technology to a technology company driven by AI technology, so that AI can move from the cost center to the core of value, and become the underlying engine for our strategic decision-making, product innovation and talent management," said Long Ling.
What is more noteworthy is that apart from external product forms, a top-down efficiency reform is underway inside Timi Tech. AI is no longer an exclusive tool of the technology department, but is defined as a "co-working partner" for every front-line business employee.
This in-depth reconstruction of the organizational foundation further reflects its open posture to embrace the AI wave. Because only when everyone can think about how to use AI to solve problems, can Timi Tech's AI ecosystem truly be full of vitality.
No matter observing the products from an external perspective, or looking at the organization's posture towards AI from an internal perspective, the "people-oriented" style and attitude seem to be indispensable for AI enterprises. In the reconstruction of real life brought by technology, Timi Tech may provide a highly valuable sample of AI evolution for internet manufacturers. Constantly pursuing certainty with actions is what makes it a bold player in the game for the brave.
Based on voice genetics, led by AI products
In 1995, the American fantasy adventure film *Jumanji* directed by Joe Johnston was released. The film tells the story of a mysterious board game "Jumanji" that will trigger dangerous events corresponding to the grid the player lands on after rolling the dice in the real world, which can only stop when someone wins the game.
Song Ke, Founder and CEO of Timi Tech, once shared in an internal speech: "In the era of rapid development of AI technology, every choice is full of unknown risks, and also brings unlimited imagination. Prosperity is the carnival for all, while adversity is the paradise for the brave. Only among those brave people who embrace AI can the leaders of the future era emerge."
For any startup, opportunities and challenges always coexist. Companies engaged in AI are like playing a mysterious game of Jumanji, choosing to roll the dice in the AI wave full of uncertainty — All in AI, All in China.
Timi Tech's attitude towards AI is not drifting with the tide, but pragmatic and rational. In terms of product form, it also thinks about how to let more people feel the charm of AI and bring technology closer to people.
For example, the Timi Tech team launched "Tunee", China's first conversational music creation agent, as well as the world's first generative AI guitar TemPolor. Its products are extremely easy to use. In Long Ling's view, the real value of AI lies in greatly lowering the technical threshold and broadening the boundary of musical expression. From models and applications to hardware, Timi Tech insists on technological and product innovation to create a more immersive creation experience for users, so that music can truly be integrated into the daily life of ordinary people.
The creativity of the AI guitar lies in that it gives everyone the possibility to create music. It is no longer a game monopolized by elites. Any young person who does not know anything about music theory can generate songs with strong emotional resonance with just a sentence or an inspiration. This is also the best presentation of a good AI product: ordinary people can quickly acquire capabilities they did not have before without arduous practice.
"We always adhere to the principle that AI is not designed to replace human beings, but to enhance human creativity and emotional expression. AI can amplify manpower by 100 times or 1000 times, but there must be a '1' in front of the multiplied number. Human imagination, aesthetic judgment and decision-making are exactly that '1'," Long Ling told 36Kr.
In terms of achievements, the GMV of TemPolor AI guitar has exceeded 10 million yuan within 3 months after its launch. Tunee has also become one of the fastest-growing products that stand out in China's AI Agent track, which also allows Timi Tech to build a solid first-mover advantage in the direction of integrated AI music software and hardware.
Making AI music easily accessible to ordinary consumers is the first step. How to make high-quality content seen by more people, enter the international market and achieve sustainable commercialization has become the next proposition of AIGC. Based on this, Timi Tech has built the world's leading one-stop video translation and AI dubbing platform — All Voice Lab, taking self-developed AI technology as the foundation to build a full-link solution integrating multilingual translation, highly emotional dubbing and overseas social media monetization for the cultural and entertainment industry.
For Timi Tech, the achievements of MaskGCT are probably the gift of the era for its "people-oriented" concept.
In the second half of 2024, Timi Tech and The Chinese University of Hong Kong (Shenzhen) jointly open-sourced the MaskGCT speech large model, which refreshed several global SOTA records. At that time, if people only focused on the technology competition, it might be difficult to tell the winner. From the perspective of industry gaps, new content forms such as Chinese micro dramas and domestic animation have gradually become popular in the global market, and Chinese culture is ushering in an unprecedented "golden period for going global", which means content localization will become a key link.
There are no dedicated AI translation and dubbing products for film, television and animation content in the market, which is exactly the opportunity for the commercial implementation of MaskGCT. Timi Tech quickly adjusted its direction and launched the AI video translation and dubbing product "All Voice Lab" within 2 to 3 months, hoping to further lower the threshold for domestic content to go global with AI technology.
For example, in the process of model training, more attention is paid to introducing knowledge such as cultural backgrounds and custom differences in different regions, so as to achieve more accurate, professional and authentic translation results. In terms of product functions and experience, through Timi Tech's previous experience in making overseas short dramas and in-depth co-creation with short drama overseas customers, it has developed a one-stop translation and dubbing workflow that truly adapts to the real business process. Within one month after launch, it reached cooperation with many leading film and television companies, and quickly received positive feedback from the market.
This creative breakthrough solves the long-standing "impossible triangle" dilemma of "quality, speed and cost" in multilingual content production faced by the overseas content distribution industry. The cost of AI translation and dubbing is 90% lower than that of manual work, the speed is increased by more than 50 times, and the monthly translation and dubbing volume reaches up to 500,000 minutes, covering more than 30 countries and regions around the world, enabling Chinese short dramas, films, animations and other content to go overseas in a large-scale and high-fidelity manner.
In terms of corporate genes, Timi Tech originated from mobile voice services, and audio technology is its inherent core barrier.
Compared with the "mechanical flavor" and "script-reading sense" of traditional speech synthesis, the root cause lies in its overly regular training corpus. Whether it is customer service dialogues or podcast recordings, the intonation of speakers is generally in a rational and stable controlled state, lacking the improvisational elements in real human communication.
Therefore, for Timi Tech, which has exclusive audio technology, choosing to develop new businesses in the vertical scenarios it excels at, digging deep and refining in the vertical field, and being a "pragmatist" that solves real problems, is undoubtedly the best springboard for development.
For Timi Tech, AI is a new business. The "newness" here refers more to the ability of continuous innovation. From the perspective of Timi Tech's team, it also has the confidence to jump into the game for the brave countless times. For example, Timi Tech's AI music team started with developing online music playing and singing apps. They explored the combination of music and AI very early, and most of the members are scarce talents who understand both music and technology.
"It is very difficult to form a team like ours in a short period of time. Of course, the underlying foundation comes from our persistence in what we love, our respect for users, and our pursuit of innovation. These things flowing in our blood cannot be copied by others," said Long Ling.
The people-oriented team construction allows Timi Tech to open a new chapter at any time. Facing AI, they do not just regard AI as an accessory or supplement, but rethink, build and operate a brand new species based on AI.
Make all team members AI players
"Cultivating an AI atmosphere cannot stop at the R&D end. It is necessary to reach a company-wide consensus: everyone embraces AI, everyone dares to use AI, and everyone is good at using AI," Long Ling said.
To embrace new changes, what is more important is the team that keeps iterating and the soil that can nurture AI. Based on this, Timi Tech upgraded the annual "Hackathon" held by the R&D team to "Maker Day" for all employees, and put forward the core proposition "What sparks can AIGC and Timi Tech create together?"
Employees form project teams on their own, gathering talents from product, operation, R&D and other positions. Driven by cutting-edge technology and interest creation, they use AIGC tools to explore how to apply the capabilities of large models to Timi Tech's business scenarios.
In addition to the iteration of internal organization, Timi Tech also fertilizes the soil from the outside.
In recent years, Timi Tech has visited major universities to carry out in-depth industry-university-research cooperation in the field of artificial intelligence, and introduce outstanding external AI talents.
Long Ling told 36Kr that in the AI era, the way of "nurturing the soil" will also change accordingly. First of all, the organizational structure can be more adaptable to local conditions. For traditional businesses, a relatively stable process-based management can be maintained. For AI-native businesses, a more flat and flexible organizational environment is needed to better adapt to the rapid development and uncertainty of AI technology.
Secondly, "embracing AI" should become the default capability of the organization, rather than an extra burden for employees. We always advocate "interest creates value", encourage employees to use AI more in their daily work, and gradually internalize it into organizational capabilities, instead of letting employees perform tasks passively. At the same time, the company will provide everyone with a better AI learning environment, allow employees to make mistakes within a reasonable range, so that everyone dares to innovate and make breakthroughs.
Facts have also proved that only when every front-line employee is thinking about how to use AI to optimize their own business, can the enterprise's AI soil truly breed a vibrant ecosystem.
For example, the exploration of AI music is actually full of difficulties, because the evaluation criteria of music are inherently subjective and diverse, and it is even more difficult to make AI fit human aesthetics. The Timi Tech team initially chose the traditional AI technical solution, using a preset framework to guide users to create, but the generated songs were like "opening a blind box", sometimes amazing and sometimes deviating from aesthetics, and never reached the ideal effect.
Over the past two years, the team has been designing different tasks, trying in multiple directions, and making choices and adjustments based on user needs. In this process, they also have to constantly deal with the matching problems of time, manpower and resources, as well as the gaps in training direction, training effect and data quality, and make timely judgments and quick adjustments.
Finally, Timi Tech successfully developed and launched the world's first multimodal music generation large model "TemPolor", and released the V4.7 version during WAIC. According to third-party tests, compared with mainstream industry models such as Suno V5.5, it scores high in indicators such as content enjoyment, work memory point and sound clarity, and other indicators of musicality, coherence and naturalness are also in the first echelon.
"Innovation is not afraid of taking detours. The key is to find the direction that truly meets user needs."
The waves change, while the original intention remains unchanged
To understand the confidence and success of Timi Tech in developing AI today, we have to turn the clock back more than ten years. The development history of Timi Tech is essentially a road of continuous innovation.
In the early years, the mobile game industry was on the eve of an outbreak. Medium and heavy competitive mobile games represented by *Honor of Kings* rose rapidly, but there was no easy-to-use real-time voice tool. Timi Tech keenly observed this seemingly small but crucial pain point, and launched TT Voice that enables smooth communication during team battles. The product attracted millions of young users in the first year after launch.
But challenges have always existed. As mobile game giants began to build built-in voice functions in their games, and large internet companies also entered this field on a large scale with their volume advantages, many people thought that TT Voice would face great challenges at that time.
Timi Tech did not over-strive for tool functions, but turned its attention to the emotions and relationships behind users. Taking voice as the starting point, it extended to the pan-entertainment social direction, and finally built a voice social ecosystem