AmazonGranted patentAudio & voiceInterest 6/10

Amazon patents playlists that auto-generate audio intros for clips

Patent drawing
Patent drawing from the official filing.

Amazon describes a system that takes short clips from long audio files and groups them by topic. It uses a large language model to write and generate unique audio introductions for each clip, which are then added to a playlist for listeners.

Why it mattersThis could help listeners discover specific parts of long podcasts or audiobooks without having to search through hours of content manually.

In the filing’s own words

Techniques are described herein for generating content segment playlists that include content segments that correspond to a sub-portion of longer content. Segments can be grouped by topic or any suitable segment attribute. Transitions to be utilized before or after playback of a content segment and/or content segment playlist may be generated using a transcription of the content segment (or multiple content segments) as input to a large language model, the output of which may be provided to a generative audio system configured to generate speech from text. The audio version of the transition may be paired with the segment or playlist to which it correspond and added to a content segment playlist. The playlist may be presented at a user interface to enable users to discover new audio content and to engage in audio content in a new way. The playlist may be subsequently updated without further user interaction.

What exactly they claim (claim 1)

1. A computer-implemented method, comprising: obtaining, by a content management system, a plurality of audio content files; generating, by the content management system from an audio content file of the plurality of audio content files, an audio content segment corresponding to a sub-portion of the audio content file; indexing, by the content management system, the audio content segment with respective metadata comprising a topic identifier that indicates a topic of the audio content segment; identifying, by the content management system, a plurality of audio content segments from a data store for a first audio content segment playlist based at least in part on executing a search query against the data store of metadata instances; generating, by the content management system, a plurality of audio introductions comprising an audio introduction for the audio content segment, a respective audio introduction of the plurality of audio introductions being generated based at least in part on generating a natural language introduction for a respective audio content segment of the plurality of audio content segments, at least a portion of the natural language introduction comprising language that differs from that provided by the audio content segment and a corresponding metadata instance; augmenting, by the content management system, the first audio content segment playlist with the plurality of audio introductions; and causing, by the content management system, presentation of the first audio content segment playlist at a user interface, the first audio content segment playlist comprising the audio content segment and the audio introduction that was generated for the audio content segment, the user interface being configured to enable playback of the first audio content segment playlist.

Facts

Company
Amazon (Amazon Technologies, Inc.)
Official title
Techniques for generating content segment playlists
Patent number
US12801664B1
Application number
18750578
Filed
Jun 21, 2024
Granted
Nov 17, 2026
Inventors
7
Claims
20

Classified under

G06F 16/639 Physics & computingG06F 40/40 Physics & computingG10L 13/08 Physics & computingG06F 3/165 Physics & computing

More from Amazon in Audio & voice

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.