AI / Software Patent Example
System and Methods to Predict Winning TV Ads, Online Videos, and Other Audiovisual Content Before Production
At a Glance
- Patent NumberUS 12,020,279
- IssuedJune 25, 2024
- FiledApril 22, 2022
- InventorTimothy Christopher D’Auria
- Claims17
About This Invention
Producing and testing video is expensive — a television campaign can cost thousands of times more to optimize than a search-engine campaign. This patent protects a system that uses artificial intelligence to predict which video concepts and attributes will succeed before any video is produced. The system analyzes existing video content across visual, audio, and metadata attributes — the number of people on screen, facial qualities, manner of speech, music mood, themes, plots, and on-screen text — measures how effective each video actually was, and then recommends the specific attributes a new video should include to meet a performance objective.
Abstract
The invention discloses a system and methods for predicting attributes of effective television ads, digital video, and other audiovisual content before production. Output comprises instructions for producing videos optimized to achieve at least one performance objective. Some embodiments predict performance of existing video, such as predicting winning Super Bowl ads before they air. The system comprises one or more processors and a memory configured to receive at least one data stream of audiovisual content in at least one public and/or a private domain; analyze said data streams to determine one or more attributes associated with said data streams; analyze said data streams to determine one or more performance scores associated with said data streams; attribute at least a portion of one said performance score to at least a portion of one said attribute associated with said data streams, and output, to a memory, any or all combinations thereof.
Representative Claim
Claim 1 of 17 — the complete claim set is available in the full patent record.
1. A method of transmitting a video between two network nodes on a packet-switched network, the method comprising:
a. at a first network node, receiving or sending a video feed via said packet-switched network and determining, for each video in said video feed, via artificial intelligence, by way of parsing the video feed and extracting elements there-within, wherein at least three of the following:
i. extracting an audio feed associated with the video feed and determining a manner of speech used by human figures within each said video further comprising associating audio with each human figure of said human figures in step ii.;
ii. extracting from said video feed quantity of human figures within each said video based on image recognition of said human figures extending through multiple frames of said video feed;
iii. determining a quality of a face of a human figure within each said video;
iv. determining a quality of an object in contact with a human figure within each said video;
v. extracting an audio feed associated with said video feed and determining a mood of music associated within each said video;
vi. extracting an audio feed associated with said video feed and determining a theme of each said video;
vii. extracting an audio feed associated with said video feed and determining a plot of each said video based on a comparison of previous plots with dialog extracted using speech to text recognition of said audio feed;
viii. determining movement of objects within each said video;
ix. extracting for the video feed alphanumeric characters displayed within each said video;
b. at a second network node, receiving or sending videos to an end-user video display device located at a third network node;
c. determining effectiveness of each said video based on one of increased revenue and increased data received by third parties as a result of said parsing of said video, by subtracting or multiplying with a baseline expectation of at least one of said increased revenue and said increased data received by third parties in relation to time since airing and/or publication of said each video in said video feed;
d. based on said determined effectiveness across a plurality of videos in said video feed, outputting a recommendation of markers to be injected into in a new video based on said determined effectiveness.
Have an invention like this?
Talk directly with a patent attorney about protecting your software and AI invention. We handle the entire process — from the first consultation through issuance — and wherever possible we work on flat fees, so you know your cost up front.