AI-Powered Video Asset Discovery For Internal Operations
Role: UX Researcher & Product Designer
Methods: User interviews, surveys, qualitative analysis, UX design, product roadmapping
Stakeholders: Marketing, Producers, Editors, Analytics teams
Goal: Help internal teams quickly discover relevant video clips using AI-powered natural language search and metadata.
I led UX research and design for an internal search tool that enables business teams to locate video moments using natural language queries. The solution reduces reliance on institutional knowledge and eliminates hours of manual video review across multiple asset management systems.
Warner Bros. Discovery business teams demand easier answers to their broad content questions. Currently, stakeholders require deep knowledge of which systems to search and what questions should be asked where. Result sets from multiple applications may need to be manually aggregated and reviewed to provide further manually narrowed lists of content. These lists may potentially need to be watched, requiring hours of content review to solve questions.
This Searchable Content effort intends to augment our ongoing MSC natural language search initiatives with a deep knowledge of content internals. By using machine learning to analyze key portions of our media library, creating valuable sets of defined time-based metadata, the tooling offers an extensible encyclopedia of content.
Background
Who will use this tool?
user research
Role
Use Case
Producer / Editor
Pinpoint specific shots to include in upcoming show or film’s teaser or trailer without scrubbing through the entire film / season
Marketing Ops
Facilitate smoother access to content and faster deliverable turnaround by ensuring the right content can be easily
retrieved by the right people
Marketing Strategist
Find high-engagement moments that can elevate new highlight reels, promos and teasers according to campaign
objectives and campaign tone
Media Team
Optimize media buying decisions by referencing content performance tied to specific audiences
Direct Marketing CRM
Retrieve customer segment-specific content moments that align to direct email, push notification, or retargeting
campaigns
Measurement Analytics
Run analyses to understand specific metadata in content that drives desired customer actions and behaviors
Agencies (External)
Enable agencies to find high-engagement moments that can elevate new highlight reels, promos and teasers
according to campaign objectives and campaign tone
Partners (External)
Allow licensing partners to browse content for promos and other monetization opportunities
Who will use this tool?
The team's Product Manager and I had touch-points with the following internal departments
user research
Role
Use Case
Producer / Editor
Pinpoint specific shots to include in upcoming show or film’s teaser or trailer without scrubbing through the entire film / season
Marketing Ops
Facilitate smoother access to content and faster deliverable turnaround by ensuring the right content can be easily
retrieved by the right people
Marketing Strategist
Find high-engagement moments that can elevate new highlight reels, promos and teasers according to campaign
objectives and campaign tone
Media Team
Optimize media buying decisions by referencing content performance tied to specific audiences
Direct Marketing CRM
Retrieve customer segment-specific content moments that align to direct email, push notification, or retargeting
campaigns
Max: 6 participants
US Networks: 8 participants
Franchises: 4 DC participants, 2 Harry Potter
Studio Tours: 5 participants
Motion Pictures: 1 participant
Teams involved
Measurement Analytics
Run analyses to understand specific metadata in content that drives desired customer actions and behaviors
Agencies (External)
Enable agencies to find high-engagement moments that can elevate new highlight reels, promos and teasers
according to campaign objectives and campaign tone
Partners (External)
Allow licensing partners to browse content for promos and other monetization opportunities
historic perspective
Providing content for the original iPhone ad - the ask required 6-10 people more than 2 weeks of research to brainstorm, 19 review, and resolve rights questions.
“Where are there clips of top tier WBD content, with WBD owned rights in the US, that contain 17 someone answering the phone and saying, ‘Hello’”
- WBD Employee
As as Marketing and Editor Specialist, I would like to perform a semantic search using natural language prompt first, then filter based on classifying and other information so that I can find at least 20-50 clips fast and efficiently, improving my searching experience.
Marketing and sales specialist
Tasks
A user types their metadata search into the universal search bar, and returns results in a grid format.
A user selects a specific clip they need further information for.
Exploring how the search results might yield ' similar search results' if the one the user searched for is not quite right. Transcript information on the right
Search
"Show me The Rock doing the eyebrow thing"
Returns:
30 relevant clips
OUR FOCUS: CONTENT METADATA
Descriptive information about the content to inform a deep understanding; can be captured at a scene, shot, sentence, or even frame level depending upon the granularity needed
Actions, subjects, objects, faces, emotion, location, etc.
EX: “The Rock wears a white t - shirt, sits on a black recliner, and stares with one eyebrow up at the camera”
media selection
Media Request Checkout Step 1: Review
Media Request Checkout Step 2: Delivery Details
Media Request Checkout Step 3: Summary
Quick export
USER FEEDBACK
01.
What Users Liked
“This saves hours of scrubbing through episodes.”
“Searching with natural language feels much more intuitive than the archive tools we currently use.”
“The similar clips feature helps us discover moments we wouldn’t have thought to search for.”
Areas for Improvement
• More advanced filtering options
• The ability to save clips to collections
• Easier ways to share clips with teams
• Greater transparency around AI metadata
Impact
• Reduce content discovery time from hours to minutes
• Enable non-technical users to search large media libraries
• Unlock new creative exploration through AI metadata
Future Opportunities
Future iterations may include:
• Campaign-based clip collections
• Editing workflow integrations
• Rights and licensing filters
• Performance analytics tied to clip attributes
02.
03.
04.