Microsoft AI can play sounds in just 3 seconds

progress Technique They can be seen in several areas: from devices that can clean the floor of your house, devices that talk, and now a robot that imitates human voices. Microsoft’s breakthrough artificial intelligence, or AI, can imitate a human voice in just three seconds.

In general, AI voices are very characteristic of robots, as it is noted that the conversation “to me“. training Amnesty International To replicate human voices, it took 60,000 hours of audio with English speaking people.

communicate AI VALL-EThe program is able to compose speeches from text and also convey feelings through what is said. The tone of voice can also be changed, in addition to the fact that the technology will use expressions that belong to human language.

Microsoft has created VALL-E, the AI ​​software that simulates voice

in portuguese Microsoft artificial intelligence is a kind of “text-to-speech” (text to speech) and was a subject of technology studies at Cornell University in New York City (USA). Basically, this technology is capable of reproducing human speech, with sound quality, in just three seconds.

Small excerpts from the reproduced text are used in different ways. The same text can be read in different ways, causing different emotions in what is being passed.

Most of the time, the AI ​​makes perfect reproduction of texts, but some specific points make the feature more robot-like. Anyway, for most of our testing, the audio was able to represent readable text well.

counterpoint at the time of testing

AI VALL-E will be able, for example, to perfectly simulate sounds without authorization. Therefore, the extension of creation can effectively become a bad thing when morality is not respected. It can easily be used for phone scams or to reproduce messages from politicians False newsFor example.

For this reason, Microsoft’s artificial intelligence is not publicly available, and only detection studies can be used.