Twelve Labs $100M Series B: Video Understanding AI Every Creator Needs to Know

Date: July 12, 2026 | Reading Time: 7 minutes

Why $100 Million for Video Understanding Matters to Creators

Twelve Labs raised $100 million in Series B funding on July 1, 2026, bringing its total funding to approximately $150 million. The round was co-led by NEA and NAVER Ventures with participation from Amazon, Radical Ventures, Index Ventures, and Red Bull Ventures. The investor quality alone signals that video understanding — not video generation — is where the infrastructure value sits.

For video creators, the distinction matters. Every AI video generation announcement this year promises faster, longer, more realistic clips. But generation is only half the workflow. The other half — finding, organizing, retrieving, and analyzing existing video — is still done primarily by humans scrubbing through timelines, manually naming clips, and maintaining sprawling folder hierarchies.

Twelve Labs’ Series B validates that the market recognizes this gap. Their models, Marengo and Pegasus, treat video as searchable, indexable data rather than a linear playback stream. For any creator with more than a few hundred hours of footage, the shift from “watch to find” to “search to find” changes how quickly content can be repurposed, analyzed, and monetized.

Marengo: Search Your Video Library Like Google Search

Marengo is a multimodal video understanding model that searches video using natural language queries. A creator describes what they are looking for in plain English — “the moment the presenter shows the Q3 growth chart and says ‘triple digit growth'” — and Marengo finds the exact frame across the entire library.

What makes Marengo different from traditional video search tools is how it understands content across multiple dimensions. It processes visual content (objects, scenes, actions on screen), spoken dialogue (transcribed and indexed in context), on-screen text (titles, captions, lower thirds), and the temporal relationships between all of these elements. A search for “the segment where the client testimonial plays after the product demo” understands the sequence — testimonial comes after demo — rather than just matching keywords.

For a creator managing a library of hundreds or thousands of hours of footage, this eliminates the most time-consuming part of video production: finding existing content to repurpose. A YouTube creator with 500 videos can search their entire back catalog for every time they mentioned a specific topic, showed a specific product, or used a specific visual treatment. The AI finds the frames, not just the videos containing them.

Pegasus: Analyze Your Video Content Automatically

Pegasus goes beyond search to reason about video content at scale. While Marengo answers “where is this moment?”, Pegasus answers “what patterns exist across my entire library?”

Pegasus can automatically generate structured metadata from unstructured video — identifying topics, speakers, visual elements, emotional tone, and content categories across every frame. It detects patterns across recordings: specific phrases that appear in multiple videos, visual elements that correlate with higher engagement, scene structures that repeat across episodes.

For creators who produce ongoing content series, Pegasus is the tool that analyzes what you have already created so you can make better decisions about what to create next. It surfaces which topics you have covered, which segments performed well, and which visual approaches you used most frequently — without anyone watching every frame.

The Business Case for Creators

The creator economy generates enormous volumes of video content. The standard workflow is: shoot, edit, publish, archive. The archive step is almost never revisited because the cost of retrieving specific content from a large library is higher than the perceived value of finding it. Twelve Labs changes that calculus.

Repurposing speed. A creator who wants to compile a “best of” compilation, create short-form clips from long-form content, or find every reference to a specific topic across years of videos can do so in seconds rather than hours. The AI finds the moments; the creator assembles them.

Content analysis without watching. Pegasus can analyze an entire video library and produce structured reports: which topics appear most frequently, which visual styles are most common, how content has evolved over time. This is market research from your own content library.

Automatic tagging. Creators who manually tag and describe their video files — for search, for licensing, for portfolio organization — can eliminate that workflow entirely. Marengo makes any frame searchable without manual labels. Pegasus generates the labels automatically.

Licensing and portfolio management. For creators who license footage or maintain client portfolios, being able to search every frame for specific content elements (brand logos, specific locations, named people) creates a new layer of value from existing assets. The content you already filmed becomes a searchable library rather than a storage cost.

Why Amazon Invested in Video Understanding

Amazon’s participation as a strategic investor in Twelve Labs’ Series B is the most revealing signal in the round. AWS is named as Twelve Labs’ preferred cloud provider, which suggests that Amazon sees video understanding as a core infrastructure layer for its cloud business — not just a niche AI application.

The integration pathway is clear: video stored in S3 can be processed through Twelve Labs’ models via API, making every frame of S3-stored video searchable without moving data to a separate platform. For AWS customers already storing large video libraries on S3, Twelve Labs becomes a native value-add that reduces the friction between storage and analysis.

For creators and small video businesses, this means the tools will likely become available through familiar infrastructure rather than requiring a separate platform. If Twelve Labs models are accessible through AWS Marketplace, SageMaker, or directly from S3, adoption goes from “evaluate a new tool” to “enable a feature on your existing storage.”

What Creators Should Do Now

Twelve Labs is currently available via API for developers and enterprise customers. For individual creators, the direct tooling is not yet consumer-ready — you need engineering support to integrate the API into your workflow. But the trajectory is clear: within 12-18 months, video understanding will be a standard feature in editing tools, content management systems, and creator platforms.

The practical step for creators today is to organize your existing footage. The biggest bottleneck to AI-powered video search is not the AI — it is getting your video into a format the AI can process. Consistent file naming, regular archival to cloud storage, and clear content categorization all make the difference between “AI can search everything” and “AI can search the 20 percent of your content that is accessible.”

When the AI tools arrive, the creators who benefit most will be the ones whose libraries are ready to search.

How Twelve Labs Compares to Other AI Video Tools

The AI video landscape has two distinct categories: generation tools that create new video and understanding tools that analyze existing video. Creators often conflate the two, but they serve fundamentally different workflows.

Runway Gen-4 Alpha — Best-in-class video generation for short marketing clips. No video understanding or search capability. Cost: $15-95/month. Use case: creating new video content from prompts.

Pika Director’s Suite — Scene-based narrative video creation with character and set persistence. No cross-video search or analysis. Cost: $10-70/month. Use case: story-driven video creation.

Meta Muse — Long-form consistent character generation for narrative content. No understanding or search layer. Use case: generating episodes of character-driven content.

Decart MirageLSD — Real-time live-stream diffusion for video transformation. No search or analysis. Use case: transforming live video feeds in real time.

Twelve Labs Marengo + Pegasus — Video understanding, search, and analysis. Does not generate video at all. Available via API; enterprise pricing. Use case: finding and analyzing content across a video library.

None of these compete with each other. They complement each other. A creator generating content with Runway or Pika also needs to organize and search the resulting library. A creator managing thousands of hours of footage with Twelve Labs also needs generation tools to create new content. The stack is: generate with one tool, manage with another.

Practical Workflow: How a Creator Uses Twelve Labs Today

Here is what a real workflow looks like with Twelve Labs integrated into a video production pipeline:

Step 1: Ingest. A creator uploads their video library to cloud storage (AWS S3 or similar) connected to Twelve Labs API. The AI processes each video, building searchable embeddings of every frame, every spoken word, every on-screen text element. Ingestion time scales with video length — a 100-hour library takes about 1-2 hours to fully index.

Step 2: Search. Instead of opening a media browser and guessing filenames, the creator types: “find every clip where I mentioned the launch date and showed the pricing page.” Marengo returns the exact frames, including timecode references for each match.

Step 3: Analyze. With Pegasus, the creator runs a library-wide analysis: which topics appear most frequently? Which visual styles dominate the first six months of content versus the last six? Which guests or collaborators appear most often? The AI produces structured reports without anyone watching the footage.

Step 4: Repurpose. The creator assembles short-form clips, compilations, or highlight reels using the search results. A “best moments” compilation that previously required days of manual review now takes minutes of search and assembly.

Every step in this workflow is currently done manually by creators with large libraries — or not done at all, leaving valuable content permanently archived and unsearchable.