TwelveLabs brings Marengo video models natively to Snowflake AI Data Cloud
![TwelveLabs brings Marengo video models natively to Snowflake AI Data Cloud](data:image/svg+xml,<svg display="block" role="presentation" viewBox="0 0 159 32" xmlns="http://www.w3.org/2000/svg"><path d="M 1.915 0 L 0.558 0 C 0.25 0 0 0.25 0 0.559 L 0 1.344 C 0 1.653 0.25 1.903 0.558 1.903 L 1.915 1.903 C 2.223 1.903 2.472 1.653 2.472 1.344 L 2.472 0.559 C 2.472 0.25 2.223 0 1.915 0 Z" fill="rgb(29, 28, 27)
Video-AI startup TwelveLabs has integrated its Marengo video understanding model natively into the Snowflake AI Data Cloud. This allows media, entertainment, and advertising teams to process video files and generate vector embeddings directly within their secure Snowflake environments, enabling analytics on unstructured video data. The integration aims to support advanced metadata queries, brand suitability scoring, and content curation for platforms like Warner Bros. Discovery, Disney, and Paramount.
Key Takeaways
- Marengo generates vector embeddings from visuals, audio, and speech via a single function call within Snowflake.
- Integrated workflows support scene-level brand suitability scoring and content curation without sending footage to external services.
- The partnership utilizes Snowflake Data Clean Rooms to enable media partners to collaborate on content-derived signals securely.
- Video processing is now treated as a standard analytics query, joining structured campaign results with unstructured video insights.
Why It Matters
Native integration transforms video from a stored cost into an actionable analytical asset by removing the friction of data movement. For major rights holders, this translates to more granular monetization via scene-level ad placement and automated metadata enrichment at scale. Within the broader ecosystem, this move reinforces the shift toward 'governed AI,' where heavy compute models are brought to the data rather than vice versa. Watch for the adoption rates of these video-native queries among Snowflake’s media customers following the Cannes Lions 2026 showcase to gauge shift in enterprise video workflows.
Additional Context
The integration of specialized AI models into cloud data warehouses reflects a broader industry trend of consolidating unstructured data workflows. Per Gartner in early 2026, enterprise interest in multimodal AI has surged as firms seek to extract value from the estimated 80% of data that remains unstructured. TwelveLabs has been a central player in this shift, having raised $50 million in Series A funding in mid-2024 from investors including Nvidia and Samsung Next to advance its vision-language foundation models. Its Marengo-V2-L model recently set benchmarks for video-to-text retrieval, outperforming previous iterations in temporal grounding and action recognition. Competitively, this move positions Snowflake against AWS Bedrock and Google Cloud’s Vertex AI, both of which have been aggressively adding specialized media models to their marketplaces. According to a May 2026 report by IDC, the media and entertainment sector's spend on AI-driven metadata extraction is projected to grow by 22% annually through 2028. By hosting Marengo natively, Snowflake aims to reduce the egress costs and security risks that previously hindered large-scale video analysis. This follow-up to TwelveLabs’ earlier deployment on Amazon Bedrock suggests a platform-agnostic strategy intended to capture the high-value media accounts that rely on multi-cloud environments for redundancy and edge processing.
Read full article at twelvelabs.io