AmazonGranted patentAI & softwareInterest 7/10

Amazon patents system to make virtual hosts demo products

This patent describes a server system that creates product demonstration videos using a virtual person. It takes a product description, writes a script, and animates a digital avatar to speak the lines and make gestures that match the words.

Why it mattersShoppers might notice these videos if Amazon uses them to replace human-made product demos with automated, consistent virtual presenters.

In the filing’s own words

Techniques are disclosed for generating a video demonstrating an item. A server device can implement an avatar video pipeline that includes one or more models for producing generative content from text input. The server device can receive an identifier for an item and obtain item information including an item description for the item. The server device can use the item information to generate a script for the video. The script can include text to be spoken in the video. The server device can use the script to generate an avatar for the video demonstration. The avatar can be animated to speak in accordance with a portion of the script and perform at least one gesture. The server device can also use the item information to generate a background scene and then generate the video that includes an image representation of the item and the avatar positioned within the background scene.

What exactly they claim (claim 1)

1. A computer-implemented method, comprising: receiving, by a server device from a user device, an item identifier for an item offered by an online retail service provider; obtaining, by the server device using the item identifier, item information comprising at least an item description; generating, by the server device and using the item information, a script for a video demonstrating the item, the script comprising text to be spoken in the video; generating, by the server device, audio from the script, the audio comprising speech corresponding to the text; receiving, by the server device from the user device, a selection of an avatar property; generating, by the server device using the script and the avatar property, an avatar for the video demonstrating the item, the avatar depicting a person characterized by at least the avatar property and animated to (i) speak in accordance with the audio and (ii) perform at least one gesture, and the at least one gesture determined based at least in part on: identifying a portion of a training video depicting a corresponding gesture and including one or more gesture words in corresponding training audio synchronized with the training video, the one or more gesture words associated with the corresponding gesture; and comparing the one or more gesture words and the text of the script to determine the at least one gesture; generating, by the server device using the item information, a background scene; and generating, by the server device, the video demonstrating the item, the video comprising an image representation of the item and the avatar positioned within the background scene, the video synchronized with the audio.

Facts

Company
Amazon (Amazon Technologies, Inc.)
Official title
Techniques for generating realistic avatar videos
Patent number
US12793306B1
Application number
18615787
Filed
Mar 25, 2024
Granted
Nov 10, 2026
Inventors
7
Claims
20

Classified under

G06T 13/205 Physics & computingG06Q 30/0623 Physics & computingG06T 13/40 Physics & computingG06T 15/04 Physics & computingG06T 2200/24 Physics & computingG06T 2210/16 Physics & computing

More from Amazon in AI & software

Summaries are simplified from the filing’s own abstract and first claim. A patent filing is not a product announcement.
Patent data and drawings: U.S. Patent and Trademark Office (USPTO). Nebula is not affiliated with or endorsed by the USPTO.