Twelve Labs is building AI that can analyze and search through videos

AI models that can understand both videos and text have the potential to unlock powerful new applications—at least, that's the belief of Jae Lee, co-founder of Twelve Labs.

While Lee may have a personal stake in this idea—since Twelve Labs specializes in training video analysis models for various use cases—there’s certainly merit to his claim.

With Twelve Labs' models, users can search for specific moments within videos, summarize clips, or even ask questions like, "When did the person in the red shirt enter the restaurant?" These capabilities are incredibly powerful, which likely explains why the company has attracted major investors, including Nvidia, Samsung, and Intel.

Video search

Video search, according to Lee, a data scientist by training, has always been flawed. While keyword searches can retrieve titles, tags, and descriptions, they fail to capture the actual content of the video itself.

"Video is the fastest-growing—and most data-heavy—medium, yet most organizations aren’t equipped to sift through their video archives," Lee explained to TechCrunch. "Even manual tagging doesn’t solve the problem. Finding a specific moment or angle in a video is like searching for a needle in a haystack."

Unable to find a suitable solution, Lee teamed up with Aiden Lee, SJ Kim, Dave Chung, and Soyoung Lee to create one. This collaboration led to the founding of Twelve Labs, a company that trains models to connect text with video content, identifying actions, objects, and background sounds.

While models like Google’s Gemini can scan video footage, and companies like Microsoft and Amazon offer video analytics services to detect objects, Lee argues that Twelve Labs sets itself apart with its customization features, allowing customers to tailor models with their own data.

“Companies like OpenAI and Google are heavily investing in general-purpose multimodal models,” Lee explained, “but these models aren’t designed specifically for video. Our distinction lies in being video-first from the start… We believe video deserves our full attention—it’s not just an add-on.”

Developers can build apps on top of Twelve Labs’ models, enabling features like video search, ad insertion, content moderation, and auto-generated highlight reels.

When I spoke to Lee last year, I raised concerns about potential bias in Twelve Labs' models—a significant risk. A 2021 study showed that training a video understanding model on local news clips, often biased in their portrayal of crime, could cause the model to develop racist patterns.

At that time, Lee mentioned that Twelve Labs planned to release model-ethics-related benchmarks and datasets. While that hasn't happened yet, Lee reassured me in our recent conversation that these tools are in development. He confirmed that Twelve Labs conducts bias testing on all its models before they are released.

“We haven’t released formal bias benchmarks yet because we want to ensure they are meaningful, practical, and actionable,” he said. “Our goal is to develop benchmarks that not only hold us accountable but also set an industry standard… Until we’ve fully achieved this—and we have a team working on it—we are committed to creating AI that empowers organizations responsibly, respects civil liberties, and drives technological progress.”

Lee also emphasized that Twelve Labs trains its models using a combination of public domain and licensed data, and does not use customer data for training.

Growth mode

Video analysis remains central to Twelve Labs' mission. However, in an effort to remain agile, the company is expanding into areas like "any-to-any" search and multimodal embeddings.

One of Twelve Labs' models, Marengo, allows searches across images and audio in addition to video, enabling users to provide a reference—whether it's an audio recording, image, or video clip—to guide the search.

The company also offers the Embed API, which creates multimodal embeddings for videos, text, images, and audio files. These embeddings are mathematical representations that capture the meanings and relationships between different data points, making them valuable for applications like anomaly detection.

With its expanding product suite, Twelve Labs has attracted clients across the enterprise, media, and entertainment industries. Two significant partners, Databricks and Snowflake, are integrating Twelve Labs' tools into their own offerings.

Databricks has created an integration that allows customers to invoke Twelve Labs' embedding service directly from their existing data pipelines. Snowflake, on the other hand, is developing connectors to Twelve Labs' models within its Cortex AI, a fully managed AI service.

“We have over 30,000 developers using our platform, ranging from individual experimenters to large enterprises incorporating our technology into their workflows,” Lee shared. “For example, we’ve partnered with municipalities on use cases like real-time threat detection, improving emergency response times, and assisting with traffic management.”

As a sign of strategic backing, both Databricks and Snowflake invested in Twelve Labs this month through their respective venture arms. They were joined by SK Telecom, Hubspot Ventures, and In-Q-Tel, a nonprofit venture capital firm based in Arlington, Virginia, that supports startups focused on U.S. intelligence capabilities.

The total new investments amount to $30 million, raising Twelve Labs' total funding to $107.1 million. Lee mentioned that the funds will be used for product development and hiring.

“We’re in a strong financial position, but we saw an opportunity to strengthen key strategic relationships with leaders who truly believe in Twelve Labs,” Lee said. “We currently have 73 full-time employees and plan to make significant investments in hiring across engineering, research, and customer-facing roles.”

New hire

Twelve Labs announced on Thursday that it is adding a president to its executive team: Yoon Kim, former CTO of SK Telecom and a key architect behind Apple’s Siri. Yoon will also take on the role of chief strategy officer, leading the company’s ambitious expansion efforts.

“While it’s uncommon for a company like Twelve Labs to hire a president at this stage, this move reflects the high demand we’ve experienced,” Lee said. He added that Yoon will split his time between the San Francisco headquarters and the company’s offices in Seoul. “Yoon is the right person to help us execute—he’ll play a crucial role in driving future growth through key acquisitions, expanding our global presence, and aligning our teams to meet ambitious goals.”

Lee also shared that the company aims to expand into new and adjacent industries, such as automotive and security, in the coming years. Given In-Q-Tel’s involvement, security (and potentially defense) work seems likely, though Lee did not confirm this directly.

“The investment from In-Q-Tel reflects the versatility and potential of our technology across many sectors, including national security,” Lee said. “We are always open to exploring opportunities where our technology can make a positive, meaningful, and responsible impact in line with our ethical guidelines.”

Text-generating AI is one thing, but AI models that can understand both images and text open up entirely new possibilities.

Take Twelve Labs, for example. The San Francisco-based startup trains AI models designed to “solve complex video-language alignment problems,” as co-founder and CEO Jae Lee puts it.

“Twelve Labs was founded to create an infrastructure for multimodal video understanding, with our first focus being semantic search—essentially ‘CTRL+F for videos,’” Lee explained in an email interview with TechCrunch. “Our vision is to help developers build programs that can see, listen, and understand the world the way we do.”

Twelve Labs’ models work to map natural language to video content, identifying actions, objects, and background sounds. This enables developers to build apps that can search videos, classify scenes, extract topics, automatically summarize content, and break videos into chapters, among other capabilities.

Lee points out that the technology can also drive applications such as ad insertion and content moderation—for example, distinguishing between violent and instructional videos that show knives. It can be used for media analytics and even to automatically generate highlight reels or create blog post headlines and tags from videos.

Given that AI models often amplify the biases in their training data, I asked Lee about the potential for bias in these models. For instance, training a video understanding model on local news clips, which often sensationalize crime in a racialized way, could lead the model to learn biased patterns.

Lee emphasized that Twelve Labs aims to meet internal bias and “fairness” metrics before releasing its models, and the company plans to release model-ethics-related benchmarks and datasets in the future. However, he had no further details to share at this time.

“Our product differs from large language models like ChatGPT because it’s specifically trained to process and understand video, integrating visual, audio, and speech components holistically,” Lee explained. “We’ve pushed the technical limits of what’s possible in video understanding.”

Google is developing a similar multimodal model for video, called MUM, which powers video recommendations across Google Search and YouTube. In addition to MUM, Google, Microsoft, and Amazon offer API-based AI services that recognize objects, places, and actions in videos, extracting detailed metadata at the frame level.

However, Lee argues that Twelve Labs stands out due to the quality of its models and its platform’s fine-tuning capabilities, allowing customers to tailor the models using their own data for “domain-specific” video analysis.

Today, Twelve Labs is unveiling Pegasus-1, a new multimodal model designed for comprehensive video analysis. Pegasus-1 can be prompted to generate detailed reports on a video or produce brief highlights with timestamps, among other tasks.

“Enterprise organizations see the potential of leveraging their vast video data for new business opportunities. But conventional video AI models, with their limited capabilities, often fall short of the complex understanding needed for most business use cases,” Lee said. “With powerful multimodal video understanding foundation models, enterprises can achieve human-level comprehension without manual analysis.”

Since launching in private beta in early May, Twelve Labs has grown its user base to 17,000 developers. The company is now working with various industries, including sports, media, e-learning, and security, with notable partners such as the NFL.

Twelve Labs is also continuing to raise funds, having closed a $10 million strategic funding round from Nvidia, Intel, and Samsung Next, bringing its total raised to $27 million.

“This investment is all about partnering with companies that can accelerate our research, product development, and distribution,” Lee said. “It’s the fuel we need to keep innovating in the field of video understanding and bring the most powerful models to our customers, no matter their use cases. We’re advancing the industry and enabling companies to do incredible things.”

Deep Dive

Searching through video content presents a host of challenges that go beyond simply transcribing speech or recognizing images. For one, traditional keyword-based search methods—often relying solely on titles, descriptions, or metadata—fail to capture the rich, semantic meaning embedded in the video's content. This results in less accurate and contextually relevant search results.

One major obstacle is the sheer volume of content. Video platforms like YouTube or Vimeo host billions of hours of video, but without effective categorization and indexing, users struggle to find specific segments that matter to them. Unlike text-based content, where keyword search is more straightforward, videos require nuanced understanding to identify and retrieve relevant segments based on user intent. Techniques like visual recognition or semantic search, which understand the content context rather than relying on metadata, are still in the developmental stages and require substantial computational resources.

Additionally, the legal and proprietary complexities of managing video content add another layer of difficulty. Platforms must navigate intellectual property concerns when indexing or transcribing video, which can limit access to certain videos or metadata. This has historically hindered the potential of video search engines.

As AI continues to advance, especially with developments in natural language processing and machine learning, the hope is that video search will evolve to be more intelligent and intuitive. However, until then, the challenges of accurately parsing, indexing, and searching video content remain a significant barrier to building truly effective systems.

Traditional methods of video search, like relying on keywords and metadata, are widely used but increasingly fall short when it comes to comprehensively analyzing video content. These methods often rely on manually entered descriptors such as titles, tags, or descriptions, which can be incomplete or misleading. While descriptive metadata, including keywords and titles, certainly makes content more discoverable, it is heavily dependent on accurate tagging. If keywords are poorly chosen or inaccurate, the search results can miss relevant videos, making the system inefficient.

Moreover, metadata can’t address the content within the video itself. Structural metadata like scene chapters might help with navigation, but they don’t allow for detailed querying based on the actual content within the video. For example, a user searching for a specific scene in a film based solely on a description like “meeting scene” might be left with results that miss the mark if the tags used are too general. This system requires a level of human intervention to ensure that tags and descriptions match the content accurately, which is neither scalable nor reliable at large volumes.

The true limitations are exposed when trying to search for more nuanced or contextual elements—emotions, objects, or even specific actions—within a video. This is where AI-driven systems, like the one Twelve Labs is developing, can make a massive difference. By analyzing the actual content through computer vision, speech-to-text, and other AI methods, such systems enable much more precise and comprehensive searches that go beyond just what’s in the metadata. Traditional methods just can't compete when it comes to extracting meaning from the actual visual and audio content of a video, especially when the metadata is either incomplete or vague.

The demand for AI-driven solutions in video analysis has grown exponentially, and for good reason. Traditional methods of video search are time-consuming and inefficient. For example, in security, investigators often have to sift through hours of footage manually, making it difficult to respond quickly to incidents. Even automated systems struggle to match the speed and accuracy needed for real-time decision-making, especially in large-scale surveillance environments.

With AI's ability to automate video tagging, recognition, and content indexing, organizations can drastically improve their ability to search and analyze videos. For instance, AI tools can instantly categorize video content based on objects, people, or actions—making it possible to retrieve relevant clips at the touch of a button. This is crucial not only for security and law enforcement but also for businesses seeking to leverage video content for marketing, training, or customer insights. AI transforms video from a passive, hard-to-navigate medium into a searchable, actionable asset, allowing organizations to make decisions faster and more effectively.

Ultimately, as AI continues to evolve, video search and analysis will become increasingly indispensable for businesses and security sectors alike. It’s not just about keeping up—it’s about staying ahead. AI isn’t the future of video content analysis; it’s the present. Organizations that embrace this technology will lead the way in both efficiency and innovation.

How Twelve Labs' AI Works

Twelve Labs is making huge strides in the AI-driven video analysis space, and their technology stands out as a game-changer in how we interact with video content. Their flagship platform uses multimodal AI to understand videos much like humans would. This means that Twelve Labs' system doesn’t just "watch" videos; it comprehends and processes them by recognizing elements such as actions, objects, speech, and text within the video. This makes video search not only faster but significantly more intuitive and accurate, allowing users to search through video content with ease, regardless of the video's length.

One of the most impressive aspects of Twelve Labs' approach is its ability to turn complex video data into vectorized representations. This opens the door to "semantic search" — an intelligent method of searching based on the meaning and context of video content, rather than relying on tags or keywords. It’s a far more powerful tool for those dealing with large video libraries or archives, such as the NFL, which has tapped Twelve Labs' AI for its massive content vault.

Their technology’s ability to understand a video scene, for example, as simple as "a man holding a pen in an office," within seconds, illustrates just how sophisticated the AI is. This is a massive leap forward from previous video search systems that only focused on metadata or basic visual recognition.

In my opinion, Twelve Labs' push into video AI is poised to revolutionize industries like media, advertising, and education. The potential to apply such advanced understanding of video to fields like content moderation, ad placement, or educational video analysis could dramatically improve efficiency and accuracy across multiple sectors.

Twelve Labs is setting a new standard in video AI with its exceptional ability to perform complex tasks like object detection, scene recognition, and speech-to-text transcription. These technologies are game-changers, offering advanced capabilities to analyze and search through videos in ways that were once thought impossible.

Object Detection is one of the standout features of Twelve Labs' AI, making it possible to identify objects within videos with remarkable accuracy. Whether it's tracking an object across frames or detecting changes in the object’s appearance, the system is flexible and powerful enough to handle even partially obscured objects. This flexibility allows for applications across industries, from security to sports analysis, by enabling automated tagging and real-time processing​.

On the other hand, Scene Recognition takes video analysis to the next level. Rather than merely identifying objects, the system can understand entire scenes, making it capable of recognizing complex scenarios and contexts. This is crucial for tasks like classifying video content or analyzing behaviors across large video datasets​.

Finally, Speech-to-Text adds another layer of functionality. By transcribing spoken words in videos, it opens the door for deeper indexing and searchability. This feature is a boon for applications in media, education, and corporate sectors where video content often includes critical spoken information that needs to be extracted for further analysis​.

However, despite these impressive capabilities, the potential of Twelve Labs' video AI is only beginning to be realized. The system's robustness will depend on its adaptability to new and unique content, and developers will need to invest in training the AI for specific use cases to achieve optimal results. Still, as it stands, Twelve Labs has created a solution that could revolutionize how we process and interact with video content.

This trio of features—object detection, scene recognition, and speech-to-text—represents the cutting edge of video AI and shows why Twelve Labs is a company to watch in the field.

AI's ability to understand and analyze video content at a deeper level has transformative potential, and Twelve Labs is at the forefront of this innovation. Unlike traditional video search methods that rely on keywords or manual tagging, Twelve Labs' approach leverages sophisticated multimodal models capable of processing both the visual and auditory components of a video. This means AI can now "watch" a video as a human would—recognizing context, emotions, and even actions that might not be immediately obvious from a transcript or a basic label.

What sets Twelve Labs apart is its ability to provide contextual video understanding. For example, AI can distinguish between a dramatic scene and an action sequence, and it can identify nuanced interactions between characters or objects across a video. This level of analysis is critical for fields like content moderation, video indexing, and even targeted advertising. The technology also enables users to search for videos based on specific events, actions, or objects, rather than just keywords or titles.

The potential applications are vast, from improving video recommendations to revolutionizing legal discovery by making video archives searchable at a level of detail never before possible. With Twelve Labs' deep understanding of video content, we are on the brink of moving past surface-level searches, towards AI that grasps and interprets video in ways that were previously reserved for human analysis. This marks a significant step forward in not only enhancing user experiences but also in unlocking a new world of opportunities for industries that rely on video data.

Applications and Use Cases

AI-driven video search, such as what Twelve Labs is building, is poised to radically reshape how businesses, content creators, and researchers interact with video content. With the ability to deeply understand video—analyzing speech, objects, actions, and more—Twelve Labs offers game-changing benefits that can streamline workflows and enhance productivity across multiple industries.

For businesses—particularly media companies and marketing agencies—AI-driven video search is invaluable. It can help identify key moments or even generate metadata across large video archives, drastically improving content management and discoverability. Marketers, for instance, can refine their advertising strategies by pinpointing specific scenes, gestures, or objects within videos, delivering much more targeted and contextually relevant ads​.

Content creators also stand to gain from this technology. Video editing and content moderation can be automated and streamlined, freeing up time for creative pursuits while ensuring that videos meet brand standards. Beyond editing, the AI's semantic understanding enhances content curation, enabling creators to easily find footage based on context rather than just keywords​.

For researchers, especially in fields like education or legal services, Twelve Labs can revolutionize video analysis. Educational institutions could leverage its capabilities to search through lectures for specific topics, while legal teams can comb through hours of video evidence with a level of precision that was previously unimaginable​. This not only accelerates research processes but also ensures higher accuracy in extracting relevant information from hours of unstructured video content.

In essence, AI-driven video search isn't just about enhancing the efficiency of video content retrieval—it's about redefining how we interact with video, making it more accessible, insightful, and powerful for businesses, creators, and researchers alike.

Twelve Labs’ video understanding AI is set to revolutionize industries that rely on large-scale video content analysis. Media, entertainment, and education sectors are particularly poised to benefit. In the media and entertainment world, this technology eliminates the cumbersome, manual process of video tagging and makes searchability much more intuitive. For content creators and streaming services, AI-powered video search will not only speed up workflows but also enable much more precise content discovery based on context and nuanced understanding of video, rather than relying on basic metadata or simple transcripts.

In education, the potential of AI-driven video analysis could transform learning experiences by making vast amounts of educational content easier to navigate. Imagine AI that can analyze video lectures, pinpoint important concepts, and cross-reference them with textbooks and other learning materials. This would not only save time but also enhance personalized learning, as students can quickly find relevant clips, concepts, or even examples, streamlining their study process.

While these industries stand to gain, they must also address challenges around data privacy and the integration of these systems into existing workflows. For example, educational institutions need to ensure that the data analyzed by AI remains secure and that the technology integrates seamlessly with traditional learning management systems.

Twelve Labs’ video understanding AI is set to revolutionize industries that rely on large-scale video content analysis. Media, entertainment, and education sectors are particularly poised to benefit. In the media and entertainment world, this technology eliminates the cumbersome, manual process of video tagging and makes searchability much more intuitive. For content creators and streaming services, AI-powered video search will not only speed up workflows but also enable much more precise content discovery based on context and nuanced understanding of video, rather than relying on basic metadata or simple transcripts.

In education, the potential of AI-driven video analysis could transform learning experiences by making vast amounts of educational content easier to navigate. Imagine AI that can analyze video lectures, pinpoint important concepts, and cross-reference them with textbooks and other learning materials. This would not only save time but also enhance personalized learning, as students can quickly find relevant clips, concepts, or even examples, streamlining their study process.

While these industries stand to gain, they must also address challenges around data privacy and the integration of these systems into existing workflows. For example, educational institutions need to ensure that the data analyzed by AI remains secure and that the technology integrates seamlessly with traditional learning management systems.

Real-world applications of AI in video content are transforming industries, and Twelve Labs is at the forefront of making videos as searchable and understandable as text. This capability allows industries to unlock the vast potential hidden within video data—traditionally a treasure trove of untapped resources. From sports analysis to content moderation, Twelve Labs' AI models enable users to not just search through video but to perform sophisticated tasks like scene classification and automated content creation.

For instance, in sports, coaches and analysts are using video to evaluate specific actions such as a swimmer’s stroke or a sprinter’s starting position, providing valuable insights to improve performance. This technology is also shaking up how media companies create content. Imagine being able to automatically generate a highlight reel of your favorite scenes, extracted from hours of footage, based on your preferences. Beyond entertainment, Twelve Labs' AI enables faster video content categorization—whether it’s surveillance footage, educational videos, or even security applications, the potential for AI-driven video classification is boundless.

In the broader context of content creation, this shift brings immense possibilities for automated video editing and the generation of summaries that are both accurate and contextually rich. Traditional video editing and searching are labor-intensive and time-consuming, but AI like Twelve Labs is changing that by offering an automated solution that breaks down video content in seconds.

Twelve Labs isn't just pushing the envelope; it's rewriting the playbook. They are tackling the challenge of understanding video the way humans naturally do—by recognizing objects, actions, and context—and turning it into an accessible format for developers and industries alike. This is a game-changer for anyone working with video content and makes us rethink how we can interact with media on a fundamental level.

The Future of Video Search

In the coming years, the landscape of video search and AI is poised to undergo a dramatic transformation, and Twelve Labs stands at the forefront of this revolution. The company's approach to multimodal AI—integrating video, audio, and contextual information—is setting the stage for a future where video content is as searchable and manageable as text. Currently, video search is a complex and fragmented process, with few solutions offering the depth and accuracy required for meaningful interactions. Twelve Labs is addressing this by making videos more accessible to both businesses and consumers, allowing them to retrieve, categorize, and understand video content with incredible precision.

As the technology matures, we can expect an explosion of applications that extend far beyond media analysis and content creation. Twelve Labs' innovative video-to-text capabilities could empower industries ranging from education to healthcare, offering new ways to process and interact with video-based knowledge. Imagine being able to search a massive library of instructional videos using simple text prompts or pulling key insights from hours of lecture content with minimal effort.

However, this progress is not without its challenges. As AI models like Twelve Labs' continue to grow more sophisticated, they will need to navigate issues such as ensuring privacy and preventing misuse of AI-generated content. Additionally, as businesses integrate these advanced tools into their workflows, the demand for real-time video search and analysis will drive further advancements in cloud-based infrastructure and computational efficiency.

Twelve Labs is setting itself up not just as a tool but as an enabler of a new way to interact with video. If they can overcome the hurdles of scaling and security, their approach could very well redefine how we think about video data—and we may one day look back at today's fragmented search capabilities as the "wild west" of video technology.

Twelve Labs is poised to significantly disrupt video content analysis with its ambitious plans to expand its AI technology. The company is focusing on refining its multimodal models, which can understand both the visual and auditory elements within videos, creating a more nuanced and scalable system for video search. One of the standout features of Twelve Labs’ technology is its ability to process video in ways that most AI tools cannot, understanding context and relationships between actions, objects, and speech. This sets it apart from existing solutions like Google’s MUM or AWS Rekognition, which primarily focus on object recognition.

Looking ahead, Twelve Labs aims to build on its foundation models for deeper video understanding, potentially revolutionizing industries that rely heavily on video content. The platform is not just about making videos searchable—it's about creating actionable insights from them. Whether it's generating automatic highlights for sports, offering precise video search capabilities for media, or improving content moderation, Twelve Labs is laying the groundwork for a future where video content is as accessible and understandable as text. Moreover, with partnerships like the one with AWS, the company is reducing costs and improving training speeds, which will allow it to scale faster and more efficiently.

However, there’s a clear challenge here. As powerful as Twelve Labs' approach is, it still faces competition from tech giants like Google and Microsoft, which already have established video recognition technologies. To truly succeed, Twelve Labs needs to continue differentiating itself by focusing on the context and quality of its AI interpretations—something it believes many larger players lack.

In my opinion, Twelve Labs’ focus on context-aware video search is revolutionary. As video content continues to grow exponentially across platforms, tools that allow for efficient, context-sensitive search will be game-changers for both creators and consumers alike. If they can maintain this focus on deep learning and understanding, Twelve Labs will undoubtedly play a central role in how video is consumed and understood in the future.

Twelve Labs is pushing the boundaries of how we interact with video content, and its potential impact on the future of video consumption cannot be overstated. By enabling users to search and analyze videos with AI that understands not just visual elements but also audio and textual cues, Twelve Labs is turning video content from something static and linear into a highly interactive and navigable medium. Imagine being able to instantly locate a specific moment in a video, or even extract detailed insights about a scene, all through a simple query. This is the future of video content, and it’s being built now.

In a world where video is fast becoming the dominant form of media—whether for entertainment, education, or business—tools like Twelve Labs are set to redefine how we engage with this content. The integration of multimodal AI models into video understanding will significantly enhance content discovery, offering more than just keyword-based searches. It will empower users to "speak" to videos, ask questions, and receive context-rich answers. This could revolutionize industries like digital marketing, where targeting and content optimization could become far more granular and intelligent.

For creators and consumers alike, this could mean a shift from passive consumption to more active interaction. Instead of scrolling through endless hours of footage to find relevant content, users could directly search for specific actions, people, or even themes, making video much more like an interactive database than a passive medium. This shift will likely pave the way for more immersive, AI-driven video experiences that could redefine everything from streaming platforms to corporate training programs.

But let’s be clear—while this is an exciting development, it also raises important questions. Will these advancements make video content even more fragmented, as viewers begin to jump from scene to scene instead of watching a video in its entirety? And how will privacy concerns evolve as AI gets better at analyzing not only what we watch but how we watch it? These are challenges that Twelve Labs, and the video industry at large, will need to address as this technology scales.

Nonetheless, the future is undeniably exciting, and Twelve Labs' technology is at the forefront of a video content revolution that will likely reshape our digital world. The impact on industries ranging from entertainment to education could be profound, making video content not only more accessible but far more dynamic. .