How AI describes VibeVoice

VibeVoice is an open-source text-to-speech platform that generates expressive, long-form, multi-speaker audio (like podcasts) from text, capable of up to 90 minutes with up to four distinct speakers. It supports cross-lingual generation between English and Mandarin Chinese and can produce context-aware emotions and singing, including background music in podcasts. The project provides MIT-licensed pre-trained models (e.g., VibeVoice-1.5B, VibeVoice-7B) and resources on GitHub and Hugging Face, with browser demos for hands-on experimentation.