Microsoft Monolinguist replicates user’s voice in translation
A new piece of translation software from Microsoft learns the tone of the user’s voice then mimics it, notes PC Pro.
Reporter Stewart Mitchell explained that the system takes about one hour of training to replicate a voice, which is then combined with a database of words; resulting in the foreign language sounding like the the person’s own voice.
It could be a sign of things to come for people keen to apply for bilingual jobs, as the Microsoft Monolinguist text-to-speech (TTS) essentially allows them to speak in another language.
The developers explained on Microsoft Research: “We have recordings of 26 languages which are used to build our text-to-speech of corresponding languages. By using our new approach, we can synthesise any mixed language pair out of the 26 languages.”
They suggested that this type of translation technology could have various uses; for instance, a person who is not completely comfortable with a language could hear a word said in a manner closer to their own speech.
Mr Mitchell pointed out that TTS software is nothing new, but what separates Monolinguist from competitors is the way it converts a person’s words into another language.
Microsoft has published a series of examples of the technology in action; sharing recordings of a user’s translations into Spanish, Mandarin and Italian.