OpenAI patents audio filter that checks if voice matches a reference
OpenAI describes a system that splits generated audio into small sections and compares each one to a specific reference voice. It uses a classifier to score how similar each section is to that reference, helping decide if the audio is safe to play.
In the filing’s own words
In one aspect, a method includes determining, by the safety engine, for a first window of the audio stream, a first score indicating a similarity between audio of the first window of the audio stream and a reference voice, where the first score is output from a classifier; and determining, for a second window of the audio stream, a second score indicating a similarity between audio of the second window of the audio stream and the reference voice, where the second score is output from the classifier, where the first score and the second score are usable in determining whether to output the first window of the audio stream and the second window of the audio stream.
What exactly they claim (claim 1)
1. A method comprising: receiving, by a safety engine, an audio stream output by a generative response engine; determining, by the safety engine, for a first window of the audio stream, a first score indicating a similarity between audio of the first window of the audio stream and a reference voice, wherein the first score is output from a classifier; and determining, for a second window of the audio stream, a second score indicating a similarity between audio of the second window of the audio stream and the reference voice, wherein the second score is output from the classifier, wherein the first score and the second score are usable in determining whether to output the first window of the audio stream and the second window of the audio stream.
Facts
- Company
- OpenAI (OpenAI OpCo, LLC.)
- Official title
- SAFETY CLASSIFIERS FOR AUDIO OUTPUT
- Publication number
- US20260341299A1
- Application number
- 19188100
- Filed
- Apr 24, 2025
- Published
- Oct 29, 2026
- Inventors
- 8
- Claims
- 20
Classified under
G10L 25/51 Physics & computingG10L 15/063 Physics & computingG10L 15/22 Physics & computingMore from OpenAI 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.
