Neosapience develops AI voice actors that sound like humans, not robots

In February 2020, a South Korean broadcasting network aired a virtual reality (VR) reunion between a grieving mother and her late daughter. The documentary was heart-wrenching and shockingly real because the VR technology had recreated the deceased daughter's exact appearance and voice.

Artificial intelligence startup Neosapience, Inc. was behind the voice creation, using samples from voice recordings that were taken before she passed away.

"With our AI technology, we create voices that are filled with emotions just like people's voices -- unlike the typical AI voices that sound like robots," said Kim Tae-su, the chief executive of Neosapience, in an interview with The Korea Economic Daily on July 6.

Founded in 2017, Neosapience can generate audio content by analyzing voice actors' recordings as well as recreate the voices of people that have passed away. Neosapience was also the world's first company to develop AI technology that recreates a person's speech into different languages.

In 2018, Neosapience shared a video of former President Donald Trump speaking in Korean. He hadn't actually spoken in Korean, but it was very convincing as it captured his unique tone and cadence. The startup used its technology to analyze Trump's speech and create an AI voice sample.

AI ACTORS BECOME BILINGUAL AND EXPRESSIVE 

The startup's core service is Typecast, an audio creation service that produces high-quality audio content based on scripts. It offers around 80 types of Korean AI voice actors and 20 English-speaking voice actors alongside a wide selection of emotions that range from sadness, happiness and anger.

Typecast has become increasingly popular among video content creators and streamers. The service has even begun to replace professional voice actors on TV programs. Also, the country's subscription-based e-book service provider Millie’s Library has adopted Neosapience's technology for its audiobooks. Typecast has over 610,000 members.

"Lately, our focus is on improving the acting ability of our AI voice actors to enhance user satisfaction," said Kim.

The startup plans to expand its English services by offering over 50 types of English voices. Recently the company has seen its overseas sales grow 20% to 30% on a monthly basis. Last year, Neosapience logged revenue of 1.2 billion won ($1.1 million).

"We plan on releasing an AI actor service, which is an evolved version of the AI voice actor service. It'll help users create videos that reflect different emotions based on their scripts," said Kim.

Write to Joo-wan Kim at kjwan@hankyung.com
Danbee Lee edited this article.