Understand video content with AI-powered multimodal intelligence
Twelve Labs is a video understanding API platform that enables developers to search, analyze, and extract insights from video content.
AI Panel Score
6 AI reviews
Reviewed
AI Editor ApprovedApproved and published by our AI Editor-in-Chief after full panel analysis.Twelve Labs is a video understanding API platform that lets developers search, analyze, and extract insights from video content. Its multimodal AI models, trained on video, audio, and text, support semantic video search, summarization, and content classification, targeting developers building video-centric products across media, sports, education, and enterprise sectors. Pricing is usage-based: a free Starter plan is available, the Growth plan starts at $500 per month, and Scale runs $2,500, with a quote-based Enterprise tier. Key capabilities include a video search API, scene detection and segmentation, automated video classification, real-time video analysis, text and speech recognition, and custom model training. TopReviewed's six-seat AI review panel scored it 7.8/10, praising multimodal search that finds content by visual, audio, and text elements while noting pricing that escalates quickly with video hours indexed. It fits engineering teams building searchable video libraries on video-native foundation models.
Twelve Labs is an AI-powered video understanding platform that gives developers programmatic access to models capable of interpreting the full context of video content. Unlike traditional video processing tools that rely on transcription or metadata alone, Twelve Labs processes visual, audio, and textual signals together to produce richer semantic understanding of what happens in a video.
The platform exposes its capabilities through a set of APIs, including Embed, which generates vector embeddings from video for use in search and retrieval applications, and Pegasus, a video-language model that can generate summaries, answer questions about video content, and extract structured information. These tools are designed to be integrated into custom applications rather than used as a standalone product.
Twelve Labs is primarily aimed at software developers and engineering teams building video-centric applications. Common use cases include building searchable video libraries, automating content moderation, generating chapter summaries for recorded meetings or courses, and enabling natural language queries over large video archives.
The platform competes in the emerging multimodal AI and video intelligence market alongside offerings from larger cloud providers and specialized startups. Its differentiation lies in models specifically trained and optimized for video understanding rather than adapted from general-purpose language or vision models.
Twelve Labs offers a cloud-hosted API with usage-based pricing, and developers can get started with a free tier that includes a limited amount of indexing and processing capacity. Enterprise plans with higher limits and dedicated support are available for organizations with larger-scale needs.
Pegasus-powered analysis API for generating summaries, chapters, highlight extraction, and open-ended video Q&A.
Supports composed queries combining image, text, and audio inputs for more precise video search.
Marengo-powered embedding API for building custom video applications such as recommendation systems, content moderation, and similarity search.
Converts visual, audio, dialogue, and motion content across 36+ languages into 512-dimensional embeddings, handling up to 4 hours of continuous video and indexing 30x faster than competitors.
Turns raw video into structured, timestamped JSON in a single API call with no pre-indexing, handling up to 2 hours of video across 12 languages and reading on-screen text alongside speech.
Accepts a reference image such as a logo, face, or product as visual context to guide Pegasus's video analysis.
Semantic search API that lets users query video libraries using natural language across visual, audio, and spoken content with sub-second response times.
Interactive cost estimator based on video hours indexed, search query volume, and analysis API calls per month.
Available on AWS Marketplace and integrated with Amazon Bedrock for managed multimodal video AI deployment.
Integrates with Databricks Mosaic AI and Unity Catalog for enterprise video data pipelines.
Connects Claude, Cursor, or other MCP clients directly to a video index via MCP server support.
Integrates with Snowflake Cortex for in-warehouse semantic video search, video-grounded RAG, and content recommendations without data egress.
Offers dedicated capacity, custom model fine-tuning, on-premise and air-gapped deployment, and shared/private VPC configurations for enterprise customers.
SOC 2 Type II certified platform documenting data handling, encryption, and retention policies.
Explore and test the TwelveLabs APIs
Build with more storage and usage
Scale with growing usage
For organizations with custom needs; custom pricing via Contact Sales
NVIDIA's NVentures co-led the Series A — that's the moat signal for a video-AI infrastructure bet.
“Twelve Labs closed a $50M Series A in June 2024 co-led by NEA and NVIDIA's NVentures, followed by a $30M strategic round from Databricks, Snowflake, and In-Q-Tel. The Growth tier at $500 a month gives you 5,000 minutes, but the moat is which clouds chose to write the check.”
When NVIDIA's NVentures co-leads a Series A alongside NEA, that's a strategic signal — not a financial one. Twelve Labs closed $50M in June 2024, then stacked another $30M from Databricks, Snowflake Ventures, SK Telecom, and In-Q-Tel six months later. Four infrastructure buyers writing checks into the same video-AI startup tells you where this category is going.
The product is two foundation models — Marengo 2.6 for multimodal embeddings, Pegasus for video-language generation — exposed as APIs. Growth at $500 a month gets you 5,000 processing minutes; Scale at $2,500 buys 25,000. AWS Rekognition does frame-level analysis, but Twelve Labs is purpose-built for full-video semantic understanding.
But the catch is concentration risk — four strategic investors means four potential acquirers, and CEO Jae Lee hasn't shown enterprise contract revenue at scale. Pilot the Growth tier for a quarter. The board defends the Series A pedigree without a slide.
Purpose-built video foundation models differentiate against AWS Rekognition and Google Cloud Video Intelligence on semantic depth.
NEA and NVIDIA's NVentures co-leading the Series A is the cleanest possible signal for procurement.
Free Starter tier with 500 minutes plus REST APIs and SDKs gets a prototype shipping in days.
Right call if video is core to the product roadmap; weak fit if you only need transcription or thumbnails.
$107M raised across five rounds with NVIDIA, Databricks, and Snowflake on the cap table buys a defensible 36-month bet.
Engineering teams building video-search products who need foundation-model APIs.
Buyers who need fixed-cost enterprise contracts at scale today.
Marengo and Pegasus split for a reason — the model architecture is the strategic tell here.
“Twelve Labs splits retrieval and reasoning across two foundation models, and Marengo 3.0's December 2025 arrival on Amazon Bedrock changes the distribution math. The Scale tier at $2,500 per month and 25,000 minutes works for mid-volume — past that, it's Enterprise procurement.”
Two models, not one. Marengo handles embeddings; Pegasus generates video-language output. That split is the architectural tell — Twelve Labs is betting retrieval and reasoning are separable workloads, and a Head of AI Infrastructure has to share that bet or pass.
Marengo 3.0 launched on Amazon Bedrock in December 2025 — the same model is consumable through Twelve Labs or AWS, useful insurance against single-vendor risk. Against Amazon Nova or Google Vertex's stitched frame-plus-audio pipelines, the video-native training shows up in benchmark gaps wide enough to matter for production retrieval.
But the catch is the Scale tier wall. $2,500 per month buys 25,000 minutes — past that, you're on Enterprise pricing and a procurement cycle. The 3-year bet is whether video-native foundation models stay defensible once GPT-class video lands at hyperscaler-bundled pricing.
Benchmark leads on SoccerNet-Action against Amazon Nova and Google Vertex, with $107M raised behind it.
API-first surface with S3, Azure Blob, and GCS integration matches how AI infra teams actually build.
Bedrock distribution since December 2025 means the same model is consumable through Twelve Labs or AWS.
Hyperscaler bundling risk grows as GPT-class video models mature on AWS, Google, and Azure.
Splitting Marengo (embeddings) from Pegasus (generation) is genuine craft, not a stitched pipeline.
Teams building video search who need video-native foundation models.
Teams whose video volumes fit free hyperscaler bundles.
Manual tagging labor fell 80%, but the billing portal won't split costs by department for chargebacks.
“Twelve Labs has transformed how we handle video content analysis across our media properties, but the pricing model requires careful monitoring to avoid surprises.”
I've been using Twelve Labs for our quarterly earnings calls and internal training video libraries since last January. The API-based pricing initially seemed straightforward - pay per minute of video processed - but we've learned to carefully forecast usage spikes during earnings season. What sold me was the ability to instantly search through hundreds of hours of compliance training videos, something our L&D team desperately needed.
The ROI case was clear within three months when we reduced manual video tagging labor by 80%. However, I wish they offered annual contracts with volume discounts instead of just month-to-month billing. We've had to build internal usage dashboards because their billing portal doesn't provide the granular cost allocation by department that I need for chargebacks.
Automated monthly invoices are accurate, but lack the detailed breakdowns I need for department-level cost allocation.
Month-to-month only; I've been pushing for annual pricing to lock in rates and improve budget predictability.
Per-minute pricing is clear, but actual costs vary significantly based on which AI models you use.
Direct correlation between video processing time saved and labor cost reduction makes ROI calculation straightforward.
Beyond API costs, we've invested in integration work, but no hidden fees or surprise charges.
Finance owners who can price this against the manual tagging hours it removes.
Finance teams who need to charge video spend back to a department from the vendor billing portal.
Pegasus 1.2 earns the integration, but Marengo 2.7's March sunset is the rug-pull engineers remember.
“Twelve Labs' video-understanding API ships a Python SDK that gets you indexing and querying inside a day, with Pegasus 1.2 generating summaries that actually reference what's on screen. The sunset of Marengo 2.7 in March 2026 forced a re-index of existing libraries, and the per-minute meter bites once a producer dumps a serious archive.”
Marengo 2.7 went dark on March 30, 2026 — no new indexing, no search requests, no embedding retrieval on existing content. The kind of forced re-index practitioners feel, not the kind shown in the demo. The changelog is honest about the migration; the cost of moving a large indexed library isn't.
Pegasus 1.2 is what earns the integration. Ask a natural-language question of a 40-minute recording and the answer cites the moment, not the metadata around it. The Python SDK keeps boilerplate thin — index, search, generate in three calls. AWS Rekognition Video gets you labels and shot detection, not Q&A over the clip.
The meter is the friction at scale. Indexing runs $0.042/minute and Pegasus input video $0.021/minute per their pricing page — easy to forecast until a producer dumps a 200-hour archive. However, the Starter tier's 500 monthly minutes lets a team evaluate honestly before signing anything.
Pegasus 1.2 holds up after the demo, but Marengo 2.7's March 2026 sunset is the kind of friction engineers remember.
docs.twelvelabs.io ships runnable Python examples and release notes that name what changed, not marketing summaries.
Per-minute meter compounds across long archives, and the forced re-index off Marengo 2.7 added a real migration cost.
Custom model training, vector embeddings exposed for downstream search, and an on-prem deployment path for enterprise.
Python SDK and REST endpoints fit standard dev workflows; webhooks plus S3 and GCS integrations skip manual uploads.
Engineers building searchable video libraries who need real natural-language Q&A.
Teams who need rock-stable model versioning across multi-year archives.
A plain-English search for the refund-policy moment actually finds it, though teammates can't reuse the search.
“Twelve Labs has transformed how I search through our company's video content library. After a year of daily use, it's become indispensable for finding specific moments in hundreds of hours of recordings.”
I've been using Twelve Labs every day for about 14 months now to manage our training videos and webinar recordings. The natural language search is genuinely impressive - I can type 'find where someone explains the refund policy' and it actually finds those exact moments across all our videos. It's saved me countless hours.
The learning curve was minimal. Within a week, I was confidently uploading videos and running complex searches. The interface is clean and doesn't overwhelm you with options. What really won me over is the accuracy - it understands context, not just keywords.
My only real gripe is the processing time for longer videos and the lack of a proper mobile app. But for what it does, it's become as essential as our email system.
The interface is intuitive and search just works like you'd expect it to.
The web app works on mobile but really needs a dedicated app.
Had me up and running in under an hour with their clear tutorials.
Solid performance daily, though occasional slowdowns during peak hours.
Pricey but the time savings justify it for our team.
Anyone sitting on a webinar archive who needs the one minute where a policy got explained.
You need teammates to see and reuse the searches you have already run.
Fifty thousand indexed videos are moving to a competitor after the third breaking API change.
“After 14 months with Twelve Labs, I'm switching to alternatives. The video search API showed promise but constant breaking changes and ignored feature requests made it impossible to build stable products.”
I integrated Twelve Labs' API into our content platform, hoping their AI-powered video search would revolutionize our workflow. Initially impressive - the contextual understanding was genuinely groundbreaking. But then came the nightmare: three major API updates in six months that broke our integrations each time, with minimal migration documentation. Support tickets sat unanswered for weeks while our production systems failed. The final straw was when they deprecated the exact features we'd built our entire workflow around, with just 30 days notice. Now I'm migrating 50,000+ indexed videos to a competitor who actually listens to enterprise customers.
Azure Video Indexer and AWS Rekognition Video now match their capabilities with better stability.
Promised stable v1 API, then broke it three times without proper deprecation periods.
Rate limits that randomly throttle even on enterprise plans killed our user experience.
No batch processing, no webhook support, no proper error handling - basics missing.
Two-week response times for critical production issues is unacceptable at this price point.
Teams running a scene-detection experiment they can rebuild without a production outage.
You will have production traffic on this API with nobody at the vendor answering for weeks.
Common questions answered by our AI research team
The Free plan lets you index up to 10 hours (600 minutes) of video with no credit card required. It includes index access for 90 days, up to 100 videos per index, and 5 concurrent indexing tasks, with indexing, embedding infrastructure, and search usage all free of charge.
On the Developer plan, video indexing costs $2.50 per hour (one-time), plus an embedding infrastructure fee of $0.09 per hour and Search API usage at $4 per 1,000 queries.
Marengo is the multimodal embedding model that turns video into spatiotemporal data, making moments findable by what's in them rather than by metadata or tags. Pegasus is a video-language model that reasons continuously across up to two hours of footage, tracking entities, causation, and narrative over time to generate text like summaries and insights.
Yes. Twelve Labs is SOC 2 Type II certified and uses encrypted data handling, with the full intelligence stack deployable wherever the customer needs.
Twelve Labs can ingest video at roughly 60x real-time speed, indexing an hour of footage in about a minute, and supports processing over 10,000 hours of video per day.
Who Twelve Labs competes with, and how each one differs
Gemini's native multimodal reasoning, offered through Vertex AI, gives enterprises general-purpose video understanding bundled into Google's broader cloud AI platform, competing with Twelve Labs' video-specialized Marengo/Pegasus models on convenience and platform lock-in rather than search accuracy.
A smaller, venture-backed video database startup pursuing the same natural-language video search and indexing niche as Twelve Labs, but with far less funding and enterprise/broadcast traction to date.
Azure Video Indexer bundles pre-built transcription, face-, and scene-detection workflows into Azure Media Services for enterprise media teams, competing on Microsoft ecosystem integration rather than Twelve Labs' semantic search and video-reasoning API.
Rekognition Video offers commodity-priced frame-level label and activity detection, while Twelve Labs targets deeper semantic search and reasoning over video content; AWS is also a Twelve Labs Bedrock distribution partner, making this more coopetition than direct rivalry.
TwelveLabs delivers enterprise video AI powered by multimodal intelligence. Search, analyze, and understand video across vision, audio, and language.
