Microsoft open sources innovative audio model VibeVoice-1.5B

This morning, Microsoft Research Institute released the innovative audio model VibeVoice-1.5B. VibeVoice-1.5B has made several significant technological breakthroughs in the field of speech: it can synthesize 90 minutes of ultra-realistic speech in one go, whereas most models can only synthesize speech for up to 60 minutes, and face challenges such as tone shifting and semantic disconnection after 30 minutes.