Discovery Agent uses natural language prompts to help producers and video teams quickly identify key scenes from media libraries for use in news, trailers, social posts or documentaries.
Discovery Agent is a new AI-driven tool designed to help producers, creatives and video teams quickly identify key scenes and moments from within their media libraries. It has been developed by Moments Lab, a specialist in video discovery software, who have identified two challenges in video content production.
One is the time and effort required to find relevant footage from large media libraries, and the other is the need to repurpose existing footage and content across multiple platforms and use-cases. Discovery Agent addresses both of these, supporting creative entertainment, digital content and media teams that need to move quickly to respond to audience demand for video content that interests them.
The Agent runs on Moments Lab’s multimodal AI indexing engine, MXT, and performs more like a personal research assistant than a search tool, suggesting content options the team may not have realised were available to them. As a result, users can be expected to gain more time to focus on editing and narrative decisions.
“Most video teams are sitting on thousands of hours of content they can’t easily access, or didn’t know they had,” said Philippe Petitpont, CEO and co-founder of Moments Lab. “We wanted to break down the barriers between producers and their content. The Discovery Agent gives users a fast, natural way to find what they need – even if they are non-technical, or new to the content. By asking for recommendations in the same way they might ask their colleagues, the engine will return precise clips from the media library for any type of editorial.”
Designed to support video search on a massive scale with fast-turnarounds, the Discovery Agent can be used to source footage for news items, trailers, social posts, recaps, documentaries, brand activations or archival projects. Users can use conversational, semantic search with follow-up prompts and refinements.
Discovery Agent brings an ability to reason and an understanding of user intent, including concepts and phrasing. It can recall previous conversations and give coherent, connected responses that preclude the need for users to repeat themselves.
Regarding output, the engine is capable of returning clips relevant to a particular moment, not just full video files. It also serves as a recommendation engine that suggests clips and concepts, potentially progressing the team’s creative process, and adds the ability to crawl the web as well as custom data sources to pull in further context and insights for video stories.
The Discovery Agent is available now in beta for Moments Lab users, and is on display at IBC2025 in Amsterdam, 12-15 September, with live demonstrations at the Moments Lab stand. The company is a participant in the 2025 IBC Accelerator Project, AI Agent Assistants for Live Production supported by ITN, BBC and Channel 4. www.momentslab.com