Baidu's new AI clones voice with a minute of audio

World
Europe

A leading Chinese technology company, Baidu, has developed an artificial intelligence (AI) algorithm that can clone human speech immediately.

The system just needs 60 seconds of voice and can change a female voice to a male one. It can also turn a British accent into an American one. There is more; the AI can learn to mimic various styles of speaking and also personalize text-to-speech to a new level.

“From a technical perspective, this is an important breakthrough showing that a complicated generative modeling problem, namely speech synthesis, can be adapted to new cases by efficiently learning only from a few examples,” Leo Zou, a member of Baidu's communications team, told Digital Trends.

“Voice cloning is expected to have significant applications in the direction of personalization in human-machine interfaces,” the researchers mentioned in a Baidu blog article.

Baidu’s Deep Voice research team last year had unveiled it could clone voices with 30 minutes of training material. Interestingly, Adobe also has a similar program named VoCo that can mimic a voice with 20 minutes of audio.

Last year, Lyrebird used neural networks to mimic voices of global leaders including, President Donald Trump and former President Barack Obama.

Recently, a research report titled “Neural Voice Cloning with a Few Samples” was released by a team of researchers, who used two approaches for voice cloning: speaker adaptation and speaker encoding.

“We demonstrate that both approaches can achieve reasonable cloning quality even with only a few cloning audios,” researchers maintained.

Innovations in voice cloning in the recent years have also raised debate on its ethical basis. After large-scale use of Photoshop for fake news, voice cloning has left many worried.

Concerns are being raised over the possible use of voice cloning for blackmailing, financial frauds and similar illegal activities.

(With inputs from agencies)