Jockey by TwelveLabs is an AI agent layer for video understanding that indexes, searches, organizes, and edits media libraries through conversational interfaces and developer APIs.
What the product does and how it is positioned
Jockey by TwelveLabs acts as an agentic layer designed for media comprehension and workflow orchestration across video and image libraries. Sitting atop the Marengo indexing model and Pegasus natural-language model, it interprets footage semantically to bypass the need for manual file tagging, folder structures, or conventional naming schemas.
The tool allows users to locate exact quotes or visual moments using natural language, automatically organize content into topic- or character-based collections, and assemble new edits directly into MP4 files. Developers can also interact with the system via APIs and SDKs to extract structured JSON metadata tied to frame-accurate timestamps.
Source-supported ways to use the product
Video editors and content creators query extensive media catalogs using descriptive phrases to locate dialogue or scenes and generate rough-cut highlight reels.
Marketing teams categorize commercial archives by format, creator, and hook to assess opening retention and detect talent or logo appearances.
Application developers define JSON schemas through the API to ingest video files and receive frame-accurate, timestamped structured data for playback deep-linking.
The documented workflow, where available
Link Jockey using supported integrations such as Claude, ChatGPT, or the TwelveLabs REST API and SDK.
Upload video and photo files directly or submit public URLs pointing to target media.
Issue natural-language prompts to retrieve specific moments, organize collections, or command the agent to assemble and export media.
Jockey functions as an operational agent layer coordinating TwelveLabs' core video foundation models. While the Marengo model generates unified multimodal embeddings across video, audio, image, and text, and Pegasus interprets media narrative and context, Jockey executes high-level tasks directed by user prompts.
Rather than limiting interaction to passive video search, the agent facilitates end-to-end media manipulation. Users can ask the agent to identify relevant footage, arrange moments chronologically or thematically, trim specific intervals with frame precision, and render out finished MP4 compilations.
Checks to run with your own material and workflow
What was checked and when
Answers based on the source-checked product record
Jockey supports video files in MP4, MOV, AVI, MKV, and WebM formats, alongside photo files in JPEG, PNG, HEIC, HEIF, WebP, BMP, and GIF formats.
Jockey does not use traditional facial recognition or build biometric databases; it groups people across media using multimodal visual, motion, and contextual cues.
Initiating a deletion through account settings permanently purges original files, generated indexes, associated metadata, and recorded usage history.
Jockey acts as an orchestration layer working over TwelveLabs' Marengo model, which indexes media by meaning, and Pegasus model, which understands video in natural language.
Users can interact with Jockey directly within Claude or ChatGPT environments, or build custom integrations using the TwelveLabs API and SDK.
