Computer games released in the last few years provide ample opportunities to fine-tune characters. Each player can create unique faces and try on any clothes so that his virtual avatar is fully consistent with his style. The only thing that is currently unavailable to fans of video games is changing its own vote, but that restriction will soon be lifted. It is expected that in games and communication services will be embedded technology Modulate, changing the tone of the voice recognition. To test it you can now.
The authors technology is Mike Pappas and Carter Huffman. The idea of creating technology to change the voice in 2015, when Huffman used the Prism app to make the photos style of famous paintings. He quickly realized that such processing can be done with audio, and a year later conducting experiments with generative-adversarial neural network, he received the first results.
That’s what people wanted for hundreds of years. Technology mentioned in science fiction and games, and that’s what we need to develop.
Mike Pappas, co-founder Modulate
To change male voice to female or to give him the voice of Darth Vader and other characters, one part of the neural network generates the sample Modulate the voice, while the second evaluates whether the tone and other features requirements. If it has the wrong height or other characteristics, the neural network creates options until a decent result.
The authors are worried that the technology will be subjected to harsh criticism, as it can be used by attackers to deceive people. Because of this, it was decided that it will be sold as a standalone application and will become part of the online games — it is hoped that this approach will not allow you to use the technology for fraudulent purposes.
Modulate can greatly diversify multiplayer — users can not only change the appearance of the characters, but voice chatting in real-time. As the technology does not respond to the words and tone of voice, it will be available in any language of the world. The authors hope that in the process of training the neural network will learn to recognize laughter and even change the focus.
To test the technology on its official website. For this you need to record a 10-second snippet of your own voice, to change his tone with the sliders and click “Apply”. To hear the modified voice by clicking on the play button.