Using Artificial Intelligence to Personalize
Video Ad Experiences

AnyClip’s team of world-class software engineers and data scientists have spent years developing proprietary technology that leverages AI to tag video content, weave content with relevant ads, and serve the right ad to the right person. These core technologies can be grouped into three robust layers:

Layer 1 - Artificial Intelligence-Driven Smart Video Tagging

Tree Female, 18-24 Vest Michael J. Fox Checkers Jeans Outdoors Watch Male, 45-55 House

AnyClip uses machine learning to create rich metadata for video discovery, personalization and monetization.

Scene Detection & Marking Algorithm

AnyClip’s algorithm marks the exact start and end-time of each scene, without altering the longer clip. When a content owner provides AnyClip with long pieces of content, this scene detection algorithm analyzes and determines sequences, scenes and shots. This later allows each tag to have an exact timestamp or a relationship with an entire “clip.”

Automatic Tagging Algorithm

Video is then processed by AnyClip’s patent-pending tagging technology. By applying image recognition and deep learning, AnyClip identifies objects, sentiment and advertising categories, recognizes brands and text, and is able to generate white or blacklists for advertisers.

External Tag Sources

AnyClip also import closed captions and metadata already associated with the content from external sources such as movie databases and sport statistics providers.


AnyClip’s WeaVo™ engine integrates with the automatic tagging algorithm and works to seamlessly integrate content with video ads:

  1. Tagging – During a campaign setup phase, AnyClip receives video assets from a client and tags them.
  2. Templates Creation – AnyClip has created dozens of ad templates. Each template includes a different ad length and content-to-ad ratio. The templates fit the most common narrative frameworks used for advertising
  3. Template Selection – A template is manually selected and the ad is automatically inserted into the template. Typically, between 1-10 parts of the ad are then cut out, leaving gaps in the ad. These gaps will soon be filled with content.
  4. Tag Match – The entire content library is searched based on tags of the pieces of the ad that were cut out and need to be replaced (often tags from the entire ad are consulted as well)
  5. Weaving – a proprietary algorithm then automatically weaves all matching content into the template, creating anywhere from 10 to hundreds of ads that blend content with ads, based on a defined template. Although the template is based on rigid second marks, the weaving is intelligent. Instead of forcing content into the ad without worrying about smooth transitions, WeaVo™ analyzes the colors, lighting and histogram of both the ad and content to determine the most appropriate place to weave.
  6. Soundtrack – Audio is not woven. A complete soundtrack is taken from the ad, from the content or from a third source, based on the template.
  7. Manual Verification – AnyClip employs a team of artistic, creative directors and video editors who manually verify and review each clip, make final retouches and sound adjustments.
  8. WeaVo Ad Tagging – based on the content tag, white/black lists are automatically generated for each ad for delivery and targeting purposes.

Layer 3 - Sense & Match Smart Recommendation Engine

Sense & Match™ is a proprietary Smart Recommendation Engine that matches three elements – an individual WeaVo ad (based on its tags), a given website, and an individual viewer.

External Data

AnyClip uses data collected by an internal DMP and external DMPs.
The four external DMPs provide the following information:

    • Websites – vertical, key words, visitors, unique visitors, etc
    • Viewers – demographic data (age, gender, household income, family size), behavioral data (democrat/republican, smoking, content preferences and history), etc.

Internal Data

AnyClip’s internal DMP collects information from a player that serves ads and content. It stores the following data about each viewer: Country-City-Zip Code, Ad Performance, Completion Rate (By quarters/seconds) Mute/unmute Pause/Play, Clicks, Shares, etc.

Through machine learning, AnyClip’s algorithm knows what tags perform well for each user on each site, and improve performance by serving more of the same or different ads. The more matching tags are found between a given ad, a website and a user, the greater a chance a given ad is served.

Contact Us