Key takeaways
- Databricks Ventures invests $30 million in Twelve Labs to enhance video intelligence for enterprises.
- Twelve Labs' multimodal models improve video reasoning for semantic search, summarization, and analysis.
- Integration with Databricks Mosaic AI streamlines video application development, reducing time and resources.
- Twelve Labs' new Marengo 2.7 model boosts video understanding using a multi-vector approach.
- Partnership aims to unlock insights from video data, addressing challenges in video search and understanding.
Databricks Ventures has invested in Twelve Labs to expand its partnership to support advanced video understanding for enterprise customers. Announced on December 16, 2024, the investment is part of a $30 million funding round, featuring participation from SK Telecom, Snowflake Ventures, HubSpot Ventures, and In-Q-Tel (IQT).
The investment follows the launch of a recent integration between Twelve Labs and Databricks Mosaic AI, which reduces development time and resource needs for video applications.
Twelve Labs, known for its innovative multimodal foundation models Marengo and Pegasus, delivers advanced video AI solutions for semantic search, summarisation, analysis, and Q&A. The company recently launched Marengo 2.7, a new video model that leverages a multi-vector approach for improved video understanding.
The collaboration allows enterprises to efficiently analyse, search, and extract insights from video data by leveraging multimodal embeddings and Databricks’ Vector Search capabilities.
Twelve Labs specialises in multimodal foundational models that focus on video reasoning. The company’s proprietary models capture relationships between visual expressions, spoken words, and overall context within videos. This enables enterprises to gain insights, monetise video libraries, and repurpose content at scale.
“While nearly 80% of the world’s data is in video format, enabling search and understanding on video data has historically been a challenging, manual and time-consuming endeavour,” Databricks said in an announcement.
“This investment deepens our longstanding partnership with Twelve Labs and opens the door for additional integration points between Databricks and Twelve Labs to help unlock the context contained within video data for a broader set of customers.”
The collaboration targets industries with video-intensive needs, such as media, sports, advertising, and security. Databricks aims to integrate Twelve Labs’ models more deeply into its platform and make them accessible within its Mosaic AI Model Serving solution.
Databricks said the integration will enable “advanced video understanding use cases, from searching video content using natural language to indexing and analysing clips and scenes based on text or images”.
Both companies expect that, as video data proliferates, the ability to analyse and extract value from it will become critical across industries. Databricks and Twelve Labs plan to deepen their integrations in the future.
