NetEase launched the "first sentence of Life" public welfare project, using AI to let hearing impaired people "speak" with their own voice.
CTOnews.com, July 21 (Xinhua) NetEase recently launched a public welfare program called "the first sentence of Life" in conjunction with the Zhejiang Provincial Welfare Foundation for the disabled. Hearing impaired people only need to upload 5-minute voice clips to generate the relevant audio model, so that they can use their own voice to "speak" again.
NetEase, a ▲ image source NetEase technology channel, said that traditional voice cloning products require input of complete speech data, which is not suitable for hearing impaired people who cannot pronounce semantics normally. Therefore, NetEase's AI Lab R & D team developed a "timbre restoration technology" by using its iSpeech technology, combined with a pre-training model and a fast fine-tuning algorithm model for a small amount of pure voice data.
At present, this "timbre recovery technology" has been launched in the audio book App. People with hearing impairment only need to open App, click the "sound recovery" function, and upload voice clips without complete semantics for 5 minutes. After a while, App can generate "their own voice". After that, people with hearing impairment can convert from text to voice and use their ideal voice to "communicate" with the outside world.
NetEase claimed that in the previous survey, more than 67% of the respondents said they were "willing to use timbre restoration technology", more than half of the respondents said they were "willing to recommend timbre restoration technology to people in need", and some respondents said that in addition to mobile App, they also hoped that timbre restoration technology would be used in public places such as buses, taxis, restaurants, supermarkets, banks and so on.
CTOnews.com learned after inquiry that, in addition to NetEase, Xiaomi had previously invited users to donate voices to people with speech disorders. Xiaomi said that in order to complete "voice customization", in addition to relevant algorithms, the most indispensable thing is a rich sound source library, so Xiaomi invites some users whose sound conditions match those with speech disorders to contribute their own voices and jointly help the development of the sound source library.
