TwelveLabs presents Rodeo, a video editing copilot by video search

TwelveLabs launches Rodeo, an AI video editing copilot using Marengo and Pegasus models to transform raw footage into first cuts via natural language search.

Rodeo is the new product from TwelveLabs, a startup specializing in multimodal video AI (image, sound, dialogue, and narrative context), designed as a copilot for editors and creators. Its premise: the difficulty lies not so much in editing as in searching through hours of rushes. The tool therefore aims to transform raw footage into a first cut in minutes rather than hours.

The principle relies on natural language search. The user formulates their request, for example, a shot where a character cries in front of a photo at sunset, and the tool semantically explores the entire library, including visuals, audio, dialogues, and context, before suggesting segments and assemblies. The creative direction remains on the human side, with the AI handling the tedious spotting.

The tool relies on TwelveLabs' in-house models: Marengo for multimodal embeddings used for search, and Pegasus, a model capable of analyzing long videos, extracting timestamped metadata from them, and grasping the causal sequence of a scene. This is the company's first consumer application, previously known for its API platform used in media, sports, and advertising.

Presented at NAB Show, Rodeo is in early access, available upon request via tryrodeo.io, and is primarily aimed at production teams managing large video volumes.