Jeff Lu, Creator of Viral AI Video Generator, Achieving $40 Million Invoiced ARR in 2 Years

Jeff (Jiajun) Lu, the founder of Akool, began his academic journey at the CAD&CG Laboratory of Zhejiang University before advancing his studies at UIUC. With over a decade of experience in product development at industry giants Apple and Google, he has accumulated deep technical expertise in areas ranging from machine learning to generative AI. Throughout his career, Jiajun has focused on leveraging technology to innovate content generation and improve production efficiency. In 2022, driven by a vision to create “the best commercial video platform,” he made the leap from large companies to entrepreneurship, founding Akool. Under his leadership, Akool rapidly established itself in the global market and has achieved an invoiced ARR of nearly $40 million.

?

In this interview with Z Potential, a leading Chinese media company focused on Gen Z audiences,, Jiajun shares insights into how Akool achieved Product-Market Fit (PMF) through its differentiated technology products, evolving from its initial startup positioning to a full product transformation. He discusses how Akool continues to expand its function matrix around video generation, delivering tailored services to enterprise clients and content creators. He also explores the transformative impact of technological developments on the future of the industry, with Akool’s long-term ambition to become a global leader in video production, benchmarking against companies like Adobe. Let’s dive into the story behind Akool.

1. From Apple/Google to Independent Entrepreneurship, Always Striving for Long-Term Impact

ZP: Welcome, Jiajun. Please introduce yourself and tell us about your journey from your academic background to founding Akool.

?

Jeff: Hello, everyone. I’m Jeff (Jiajun) Lu, the founder of Akool. I’ve always had a deep interest in physics and mathematics since my school years. My curiosity about business was sparked by reading “The Great Defeat” by Wu Xiaobo in high school, which led me to pursue computer science, seeing it as a perfect blend of physics, math, and business. In 2009, I began studying computer graphics at Zhejiang University, a leading global lab in the field, where I focused on AI-enhanced 3D modeling. This marked my first exposure to AIGC, and I’ve been exploring the intersection of graphics, CG, and visual generation ever since.

After my undergraduate studies, I pursued a Ph.D. at UIUC under a renowned professor in computer vision, where I worked on projects involving video editing, real-time sensor processing, and the fusion of machine learning with vision. I also co-founded Xiaohei Technology (now Today’s Watermark Camera), and created “Double Date,” a social app.

After graduating in 2018, I joined Apple and participated in the development of the Face ID project, especially focusing on face recognition when wearing masks. In 2020, I joined Google Cloud, specializing in video processing and human motion recognition, such as counting heads, analyzing human behavior and actions, and enabling more efficient retrieval and analysis of video content through video structure processing. During that time, I also served as an adviser for Xreal for a year, building the AI/ML team and jointly promoting the development of some AI/ML applications, with a core direction of gesture recognition for AR glasses control.

ZP: In 2022, you left large companies to start your own business. What motivated that decision?

?

Jeff: Entrepreneurship has always been a strong aspiration of mine. I’ve been deeply influenced by books like those by Wu Xiaobo and by the stories of figures like Bill Gates and Elon Musk. Their journeys have shaped my desire to create something meaningful. I even wrote a book called Enhanced Humanity, which explores how technology can drive human evolution—covering topics like human-machine integration, memory uploading, and even immortality. These frontier fields captivate me, which is why I’m so focused on the virtual human domain today. While Elon Musk aims to enable human survival on other planets, my focus is on transforming humanity here on Earth—making us smarter and stronger through technology. It might sound a bit sci-fi, but my goal is to create a significant impact, not just for myself and Akool but for human development as a whole. I felt staying in a large company wasn’t aligned with these aspirations. Additionally, my participation in Harvard Business School’s PLD program reinforced the idea of making bold changes. Although I had experience with two previous startups, I realized that launching my own business was the best way to pursue what truly inspired me.

ZP: How did your experience at large companies and previous startups prepare you for this entrepreneurial journey?

?

Jeff: Resilience is key in entrepreneurship, and it’s essential to align with the times. My time in large companies gave me invaluable systematic methodologies, such as understanding the full development process, building infrastructure, and creating organizational frameworks. At Apple, I learned the importance of refining products until they reached the highest possible user experience. At Google, I was exposed to cutting-edge technological innovations and how they could be applied across various industries to make a meaningful impact.

ZP: Akool’s headquarters is based in the U.S., and the founding team has a diverse international background. Can you share how the founding team came together?

?

Jeff: From the outset, I wanted to create a truly international company, which required building an equally international team. This was especially crucial given our focus on the B2B market—having the right mix of global experience and talent was essential to successfully engage enterprise customers. In the early stages, the team was built primarily through personal connections and referrals. One key individual was Deepa Sureka, whom I met in the startup community before launching Akool. We had initially discussed joining each other’s projects, but eventually, Deepa joined Akool and brought with her a network of talented colleagues from diverse international backgrounds. Over time, we expanded our recruitment efforts, reaching out to candidates via platforms like LinkedIn.

2. Benchmarking Against Adobe: Positioning Akool as the Next-Generation AI Video Platform and Achieving Product-Market Fit by Focusing on Initial Customer Needs

ZP: What was Akool’s initial positioning when it was founded, and how has it evolved to where it is now?

?

Jeff: Initially, we were exploring new opportunities within AI and saw a gap in the market for a business model like SenseTime in the U.S. Our vision was to create something similar to “SenseTime of the U.S.” From day one, we focused on the B2B market, starting with the Saudi government. Our first projects were centered around security, like face recognition-based access control, primarily targeting the Middle East and U.S. markets. However, we encountered some challenges. Government projects in the Middle East were slow, often delayed by six months or more, and the U.S. market for security technology also moved at a slower pace.

While we were initially focused on security, we launched additional features, including video editing and generation functions. By the fourth quarter of 2022, we noticed significant growth in video editing. Even without any marketing, usage increased organically, which signaled to us that this was where we were seeing the real Product-Market Fit (PMF). From there, we decided to double down in this direction, and after the release of ChatGPT, we fully pivoted to focus on this area, enhancing the product’s capabilities and refining the user experience.

?

ZP: What triggered this pivot to focus on video generation?

?

Jeff: There were two main factors. First, we saw a shift in technology, with models like Stable Diffusion maturing. While language models were evolving quickly, we also noticed that image and video technology were beginning to catch up. At that time, tools like Midjourney were emerging but still in their infancy, and we recognized that the video space was still largely untapped, presenting enormous potential. We also saw the growth in data usage from our video functions, which confirmed that this was the area where we could make a significant impact, leading us to fully commit to this direction.

?

ZP: You mentioned Akool’s positioning as a B2B Video Cloud. What were the initial products you launched, and how has the product evolved since?

?

Jeff: When we launched, we focused on three core products: Face Swap, 3D Generation, and Image Generation. While our 3D and image generation capabilities were solid technically, the product experience needed further refinement. At that time, Midjourney was gaining popularity for image generation, and we initially targeted the e-commerce sector for 3D generation. However, we later realized that demand for 3D models was weaker than expected. Face Swap, on the other hand, became our first successful product, securing a major order from Coca-Cola in January 2023, which helped build momentum.

Our product expansion followed two key strategies: First, after securing our initial customer base, we explored additional needs within that group. Second, we expanded on technologies similar to Face Swap, leveraging our existing capabilities. This led to the development of new features focused on “human” aspects, such as Face Re-editing, Reanimation, and even Video Translation. As we honed in on our core audience of marketers, advertisers, and content creators, we found that human-related editing and generation were the most in-demand features. While we also launched image and video background replacement functions, the primary traction came from human video generation and editing. Based on these insights, we concentrated more R&D resources on developing human-focused video generation tools. Overall, our product evolution was driven by the synergy between technology, customer needs, and market trends.

ZP: Why did we choose to focus on the “human” direction?

?

Jeff: There are several reasons why we chose to focus on “human” technologies, especially face-related features. First, technologies in this area, including face generation, are more mature compared to generating random objects or abstract content. Generating faces is a more controlled process, making it less complex and more feasible. Even full-body generation, while more challenging, is still within the realm of possibility and easier to develop.

Second, users are more willing to pay for products that involve human generation and editing, as the demand is higher, and there’s a stronger overall willingness to invest in these types of tools. People value personalization, especially when it comes to their likeness, which drives greater payment potential.

Third, generating objects is far more difficult. E-commerce is the primary market for object generation, but the challenge lies in creating objects that are both highly realistic and consistent with their real-life counterparts, which is technically demanding. Furthermore, e-commerce businesses often don’t have a strong incentive to create such materials themselves, as they usually source materials directly from upstream suppliers, many of which are based in Asia. In fact, these businesses often rely on contract manufacturing, where they don’t need to generate their own content.

In contrast, the marketing sector has a much stronger demand for human-centric content, coupled with a higher willingness to pay for it. Given the technical feasibility and market demand, focusing on the “human” track made the most sense.

ZP: Akool has focused on “human” video generation. What products and features do we currently offer?

?

Jeff: Our product lineup includes several key features focused on human-centric video generation. These are:

1. Face Swap – both real-time and non-real-time face swapping capabilities.
2. Talking Avatar – this allows us to generate videos of virtual characters, edit content from existing videos (such as changing what people in the video say), and support real-time interactions, like live conversations with customer support agents.
3. Video Translation – this feature translates the language in a video to multiple target languages while preserving the natural performance of the original video.
4. Human Generation Features – including Image Generation and Image-to-Image, which generate virtual characters with different appearances.
5. 3D Motion Editing – a recent addition to our product offering.

ZP: How did Akool gradually find its Product-Market Fit (PMF)?

?

Jeff: In terms of product development, we launched several features and paid close attention to market trends and user needs. For example, Face Swap quickly gained traction as a highly utilized feature, so we decided to base our product on this technology.

Regarding the market, we initially saw interest from various virtual human companies. A large Canadian virtual human company particularly stood out, advising us to pivot to avatar technology, as they saw the potential for high profitability. By the second and third quarters of 2022, many virtual human companies approached us for lip-syncing technology, which we began developing in response.?

To explore application scenarios, we tested hypotheses by sharing them in forums and communities. As a result, many users tried our product. In January 2023, we signed Coca-Cola as a client, using this partnership to reach out to potential customers. Through targeted marketing, we identified new leads, ultimately narrowing our focus to the marketing scenario.

ZP: How do we determine if we have achieved PMF?

?

Jeff: Achieving PMF is marked by unmistakably strong signals. After we shifted our focus to video production, the demand surged to the point where our servers were nearly overwhelmed, and even adding more machines couldn’t keep up. If you’re unsure whether you’ve reached PMF, chances are you haven’t yet.

ZP: Noticing that many players in this space are primarily targeting prosumers (ToP), why did Akool choose the B2B model?

?

Jeff: We analyzed the US productivity software market and observed key differences between ToP and ToB products. ToP products are lighter, have shorter decision cycles, and tend to experience faster initial growth. However, ToB products, while heavier and requiring longer refinement, offer stronger customer retention and greater growth potential in the long run. Nearly all the productivity software companies valued in the billions are ToB. Additionally, our team has a natural affinity for the B2B space and a deeper understanding of the needs of enterprise clients. From the start, our branding was geared toward B2B, initially positioning ourselves as the “SenseTime of the US,” and working in sectors like security, cloud-based AI services, and marketing. Though we’ve shifted focus over time, our core B2B branding has remained consistent.

That said, we are considering creating a separate landing page to distinguish our B2B and ToP offerings. Currently, over 60% of our customers are businesses, and about 40% are prosumers. We haven’t implemented distinct marketing strategies for prosumers yet, as we see significant growth potential in this segment. Given that Adobe targets both B2B and ToP audiences with different landing pages, we might adopt a similar approach to stay competitive.

ZP: What do you think about long-term competition with Adobe?

?

Jeff: We’re primarily focused on the “human” aspect, where Adobe has done relatively little. At the same time, we’re collaborating with Adobe by providing them with products and services. Overall, based on our work with them, we see that Adobe’s progress in the “human” space is slow. They have a lot of ideas but are cautious in execution, presenting us with a significant opportunity. Their current strategy appears to be integrating third-party solutions, and they aren’t as aggressive as startups in this area.

ZP: What new products does Akool plan to launch in the next 2-3 years? What is the logic behind expanding the product line?

?

Jeff: From a functional perspective, we will focus on expanding “human” video production capabilities, including enhancements in appearance, voice, motion, and potentially even adding elements like clothing and accessories. Additionally, we plan to explore more generic video generation as technological advancements allow, but we will proceed cautiously in expanding into broader video editing functions.

ZP: What is Akool’s long-term vision?

?

Jeff: In the long run, we aim to become a company on par with Adobe. Our goal in the coming years is to establish ourselves as the best commercial video production platform, enabling a wide range of commercial applications, from advertising and marketing to film production.

3. Targeting the Enterprise Market with Higher Ceilings and Deeper Barriers, Solving Systematic Engineering Problems in Video Generation

ZP: How do you view the current competitors in the industry? What are our differentiated advantages?

?

Jeff: Our competitors in the industry include Synthesia, HeyGen, Runway, Captions, and others, all of whom have done well in the startup space. However, in the long term, we see industry giants like Adobe and Canva as the real challengers. If Google Cloud enters the video generation space in the future, they could also become significant competitors.

Our differentiated advantage lies in our strong focus on the enterprise market. While many competitors target the prosumer market, we believe the enterprise sector offers a higher ceiling, deeper barriers to entry, and stronger customer loyalty. Enterprise customer needs are often more complex and varied, which is why we are focusing on specific verticals to provide feature differentiation. Our team has a deep understanding of these enterprise needs, particularly in video production, and this gives us a distinct edge.

Additionally, video production and editing are systematic engineering challenges. To implement video generation in diverse scenarios, a substantial amount of engineering work is required. It also takes continuous product refinement to enhance the user experience. These are areas where our team excels, making us well-equipped to solve the engineering problems associated with video generation at scale.

ZP: The company has achieved good results in commercialization. How will it continue to maintain its first-mover advantage?

?

Jeff: To date, our invoiced ARR has reached $40 million, and we got an early start in the market. Our key advantage lies in the customers and data we’ve accumulated. We’ve established a network effect among our customers by creating workspaces, collecting numerous customer cases, and refining the product experience based on their feedback. This ongoing interaction helps us continuously improve.

The network effect is evident in two main areas: workspace/collaboration and data. We are also innovating in IP and have built exclusive partnerships that will help us expand our intellectual property further. By leveraging these advantages, we’re positioning ourselves to maintain a leadership position in the market.

ZP: What key changes and breakthroughs have occurred in the technology of video production/generation and editing in the past few years?

?

Jeff: When I first started in this field, video production technology was in its infancy, and generated videos were not usable. At that time, many of the companies now rapidly raising funds were not focused on AI video generation—they were primarily web-based video editors that later pivoted to AI video generation. Back then, AI video generation was so underdeveloped that companies purely dedicated to it were very small, and larger players came from the AI video editing space.

However, there have been many significant advancements and breakthroughs in recent years. Video quality has improved drastically, and digital humans, once stiff and unrealistic, are now nearly indistinguishable from real people. A major technological advancement in the video generation field has been the introduction of Vision Transformers (ViT) for overall video generation. From an application standpoint, video generation was not widely used in the past, but today, many companies have made breakthroughs in its application, with technologies moving into full-scale production.

Overall, we’ve seen improvements in technology maturity, video effects, application reach, and model size and structure. One of the most impactful breakthroughs for our business is the shift from 2D-based algorithms to 3D space, enhancing realism. Additionally, the optimization of model structures and the increase in model size have had significant implications for our business growth and capabilities.

ZP: What new technological breakthroughs do you hope will have a significant impact on our business in the next three to five years?

?

Jeff: There are still many areas that can be optimized, and a lot of work remains to be done before we achieve perfection. Video generation is challenging. For example, simulating physical behaviors and adhering to physical laws in video creation is still difficult. Currently, most advancements rely on big data and memorization techniques, but we’re far from achieving the level of sophistication needed for video generation and editing that adheres to realistic physical laws like in movies.

This leaves ample room for innovation. For instance, fully replacing the head—including hair—in face swapping is still a complex process that requires a lot of interaction. It’s also technically difficult to generate and edit videos under different lighting conditions and angles. Moreover, while generic video generation has improved rapidly, it still faces significant limitations. Even specific cases remain a challenge. At the moment, many of these technologies are in the entertainment phase, where lower quality results are more acceptable. However, to enter full-scale production, the standards are much higher. There’s a great deal of room for improvement before we can fully scale these technologies for commercial applications.

ZP: What were your expectations for yourself 10 years ago? Have they been achieved? Standing today, what kind of person do you hope to become 10 years from now?

?

Jeff: Ten years ago, I was involved in my first startup, with the goal of building a highly successful company. While we’re still on our journey, we’re working toward making Akool a truly impactful and successful business.

Looking ahead, I hope that in 10 years, Akool will be a globally recognized company, serving millions of customers. If all goes well, I might explore new frontiers, just as Elon Musk has in space, perhaps diving into areas like “social impact” or “human creation.” I wrote a book called Enhanced Humanity, which explores how technology can drive human progress, touching on topics like human-machine integration, memory uploading, and even immortality. These frontier fields fascinate me, which is why I’m focusing on the virtual human industry. While Musk aims to enable humans to live on other planets, my focus is more on transforming humanity itself, making us smarter and stronger through technology. It might sound like science fiction, but I’m driven by the desire to make a significant impact both on a personal level and for humanity’s development. For me, staying within a large company isn’t central to achieving these goals.

?

ZP: Besides entrepreneurship, what are your hobbies?

?

Jeff: My biggest hobby is watching interesting videos. I also enjoy exploring good food and staying active through exercise.

?

ZP: Who is your favorite entrepreneur?

?

Jeff: My favorite entrepreneur is Elon Musk. He’s pursuing meaningful projects that have the potential to change human development. Entrepreneurship is an ongoing pursuit—after achieving one milestone, there’s always the next challenge. What excites me most is how to create things that can make a lasting impact on history and humanity.

For more information about Akool, visit https://akool.com/.