VibeVoice

VibeVoice is a novel framework designed for generating expressive, long-form, multi-speaker conversational audio, such as podcasts, from text. It addresses significant challenges in traditional Text-to-Speech (TTS) systems, particularly in scalability, speaker consistency, and natural turn-taking. VibeVoice employs a next-token diffusion framework, leveraging a Large Language Model (LLM) to understand textual context and dialogue flow, and can synthesize speech up to 90 minutes long with up to 4 distinct speakers.

Users expressed mixed feelings about VibeVoice's text-to-speech capabilities, noting that "the intonation is off on almost every phrase." While some praised the model's advancements, others found the voices to still sound robotic and lacking in naturalness.

Another recurring theme in the discussion was the disparity in quality between male and female voices. One user commented, "the male voices seem much worse than the female voices," highlighting a potential area for improvement in the model's design.