S. Korea's top web portal introduces new voices synthesis technology

By Park Sae-jin Posted : July 4, 2018, 18:40 Updated : July 4, 2018, 18:40

[Iclickart]


SEOUL -- Naver, South Korea's top web portal service, boasted about its new AI-based voice synthesis technology claiming it requires a smaller number of sample voices than that of Google's.

The voice synthesis technology is seen as vital for the Fourth Industrial Revolution along with AI and big data. Google introduced Duplex, an AI-based voice synthesis engine which can carry out natural conversations by mimicking the human voice.

"The problem we face with the personified voice synthesizing is that we require a lot of voice samples," a Naver official told reporters at a forum in Seoul on Wednesday. The official claimed Naver's new AI voice synthesis engine requires only about four-hour-long voice samples while Google needs at least 40-hour-long samples.

Naver said a book-reading service for children using the voices of parents or celebrities can be provided by using its new voice technology. It said its AI has learned language abilities by studying 25 million web documents stored in the company's database.

According to Naver, its AI assistant speakers can recognize the voices of individual users during the latter half of this year. 
기사 이미지 확대 보기
닫기