VibeVoice: Microsoft’s Long-Form Conversational AI Audio

Here’s a breakdown of the key information from the provided text:

VibeVoice: A New Voice AI Model

What it is: A text-to-speech model developed by Microsoft, positioned as a competitor to Google’s NotebookLM.
Key Features:
Generates up to four distinct voices.
Can produce up to 90 minutes of podcast-quality speech.
Reads and organizes text – designed to perform text audibly, like replacing a recording studio, rather than understanding it.
Runs on 1.5 billion parameters.
Uses Alibaba’s open-source Qwen2.5 large language model for natural dialog.
How it differs from NotebookLM: NotebookLM can do two voices, while VibeVoice can do four. NotebookLM ingests documents and creates podcasts, while VibeVoice focuses on audibly reading and organizing text.Voice AI Market Trends

Investment: Voice AI startups raised $2.1 billion in 2024,an eightfold increase from the previous year. Voice Shopping: Increasing adoption of voice shopping, particularly among Gen Z (30.4% shop by voice weekly) and Millennials. The average across all ages is 17.9%.

potential Applications: The text mentions potential research applications, but doesn’t elaborate further.