Baidu successfully developed AI can imitate the voice after only a few seconds of listening

The AI ​​team has successfully developed a neural network capable of mimicking voices in less than a minute. Even, this software can change the voice to another gender or accentuate the accent.

Baidu search giant, the Chinese company Google has announced a new achievement in the field. According to Futurism, the company has successfully developed a program, a neural network application capable of mimicking the voice after listening to a clip for a few seconds.

Baidu has just announced a new achievement in the field of artificial intelligence.

Not only is it possible to imitate the voice, the program can also change that voice to another gender or even accentuate the accent like a human. Readers can listen to some examples created from this program here.

Previous attempts of this technology allowed the system to replicate the voice after analyzing a longer speech pattern. In 2017, the Baidu Deep Voice team introduced technology that could duplicate the voice within 30 minutes of training.

Adobe also has a program called Voco with the ability to imitate the voice with only 20 minutes of listening. Or a start-up in Canada created Lyrebird with the ability to imitate the voice within just 1 minute. But Baidu even goes so far as successful development of AI can imitate the voice in just a few seconds.

Baidu's new AI technology is expected to help create more intelligent virtual assistants or more natural-sounding voice translation services. However, like many other technologies, voice imitation is also at risk of abuse if not controlled well.

Baidu's new AI technology is expected to help create intelligent virtual assistants.

According to New Scientist, Baidu's voice from the new AI program can fool others with 95% accuracy. People even evaluated the ability of this AI to mimic the voice with a score of 3.16 on a scale of 4. Thus, in addition to the extremely positive benefits, this AI has the potential to be exploited into very bad purposes.

Existing programs can use AI to replace or swap, even recreate from the beginning of an individual's face in a video.

For example, an AI program of researchers at the University. Washington has allowed the creation of a fake video, simulating the speech of former President Barack Obama. The AI ​​plays a role in accurately modeling Obama's mouth movement while he talks. Then, with voice-over techniques, they can control Obama to "fake" whatever they want.

The fake video simulating the speech of President Barack Obama.

Not long after the American scientists' technology raised a major concern regarding fake videos, the appearance of the program was able to mimic Baidu's voice and continue to make many people more worried. Because the situation of fake news is likely to appear more widespread in the future.