Shulou(Shulou.com)11/24 Report--

CTOnews.com July 27 news, Microsoft recently launched a speech model called NaturalSpeech2, the model uses "potential diffusion" design, at the zero sample Text To Speech level outstanding results, Microsoft claims that the model provides a "commercial grade" speech/singing solution, can give users a high quality, diverse Text To Speech experience.

Microsoft ran a series of demos of NaturalSpeech2, demonstrating its ability to generate speech with different speaker identities, prosody, and styles (such as singing) with zero samples.

It is reported that, unlike traditional speech-to-text (TTS) systems, Microsoft's NaturalSpeech2 uses "continuous vectors" instead of "discrete markers" to represent speech, thus generating more complete speech fragments without the phenomenon of "stick reading (speaking word by word)" with "lack of emotion."

The experimental results of NaturalSpeech 2 show that the prosody of speech generated by NaturalSpeech2 under zero sample condition is almost consistent with that of speech prompt and real speech, and the naturalness (measured by CMOS) on LibriTTS and VCTK test sets is difficult to distinguish from real speech.

The project's paper has been published on GitHub, and interested CTOnews.com partners can visit it here.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.