Naturalspeech model synthetic speech achieves human speech level for the first time in CMOS test

AI Synthetic speech is now common , However, it doesn't make people feel immersed like talking with real people and reading . however , Microsoft Asia Research Institute and Microsoft Azure The speech team recently jointly launched a new end-to-end speech synthesis model NaturalSpeech, stay CMOS For the first time in the test, it reached the level of human speech . This is a step closer to improving Microsoft Azure Level of synthetic speech in , Make all synthetic sounds lifelike .

Thesis link ：https://arxiv.org/pdf/2205.04421.pdf