In brief: There's been a lot of talk about AIs being able to perfectly imitate someone's voice and the potential dangers that could bring, but Samsung wants to use the technology in a less nefarious way. The company's Bixby Text Call feature, which converts text to speech for phone calls, is getting an update, so the voice it uses is an exact copy of the user's.
Bixby Text Call, part of Samsung's One UI 5.1, gives users the option of typing out a message instead of answering an incoming phone call. Rather than sending this message as a text, Bixby converts it to voice-generated audio, which it plays upon answering the call.
The service uses several default voices for reading messages out aloud. But a new update uses a replica of your own voice for those text replies. This is achieved by recording several sentences of your voice in the Bixby Voice Creator, which creates an AI-generated copy right down to the tone.
The Bixby Text Call feature is available now in English. The AI voice-generator is only available in Korea at the moment, but it is rolling out to English speakers with the Galaxy S23 series later this month. Samsung says it plans to make the generated voice compatible with other Samsung apps "beyond phone calls."
Imitating someone's voice is always going to raise fears about possible scams or other crimes should someone steal the phone and gain access to it, but Samsung notes that the feature does inform callers that there are speaking to Bixby and not the person they rang.
It was only last week that we heard of an Apple patent that could allow an iMessage to be read aloud in the sender's own voice. It sounds very similar to what Samsung offers, though Apple's version is related only to its iMessage service.
We're just two months into 2023, and it already looks like it'll be remembered as the year AI went mainstream. One just has to look at all the stories about ChatGPT, voice-generating systems, and autonomous military vehicles for an idea of what the future holds.