Microsoft is developing MAI-Realtime, a bidirectional voice model currently in internal preview, as part of its own proprietary voice AI stack. The model processes audio in real time in both directions simultaneously, enabling more natural interactions across multiple languages. Microsoft has also launched MAI-Voice-1 for expressive speech synthesis and MAI-Transcribe-1 for speech recognition covering 25 languages, creating an end-to-end vertically integrated voice AI stack. This strategy aims to reduce dependence on OpenAI and give Microsoft full control over its technology roadmap and pricing decisions. The potential integration of MAI-Realtime into Copilot for Microsoft 365 would reach enterprise users across Word, Excel, Teams, and Outlook.
Source: Read the original article
