Microsoft Debuts First In-House AI Models: MAI-Voice-1 For Ultra-Fast Speech And MAI-1-Preview For Instruction-Following Tasks

Microsoft AI, a division devoted to creating and integrating AI applied sciences throughout Microsoft, introduced the discharge of MAI-Voice-1, its first high-fidelity and expressive speech era mannequin. The mannequin is at present out there in Copilot Each day and Podcasts, in addition to within the new Copilot Labs expertise, permitting customers to discover expressive speech and storytelling capabilities. 

MAI-Voice-1 delivers pure audio in each single- and multi-speaker situations and is designed for pace, producing a full minute of speech in below one second on a single GPU, making it one of the vital environment friendly speech era methods at present out there. The mannequin permits purposes reminiscent of interactive “select your individual journey” tales or customized guided meditations, showcasing the potential of voice as a main interface for AI companions.

Microsoft AI Launches Public Testing Of MAI-1-Preview, Its First Absolutely Skilled Basis Mannequin

Aside from that, Microsoft AI has initiated public testing of MAI-1-preview on LMArena, a extensively used platform for neighborhood mannequin analysis. This marks the division’s first totally educated basis mannequin and offers an early take a look at capabilities that shall be built-in into Copilot. The MAI-1-preview is an in-house mixture-of-experts mannequin, pre- and post-trained on roughly 15,000 NVIDIA H100 GPUs, designed to deal with instruction-following and ship useful responses for on a regular basis duties.

The mannequin shall be steadily launched for chosen text-based use circumstances in Copilot, permitting Microsoft AI to assemble suggestions and refine efficiency. The staff combines in-house fashions, accomplice contributions, and open-source improvements to optimize outcomes throughout thousands and thousands of interactions day by day. MAI-1-preview can be out there to trusted testers, with API entry purposes open to gather insights on its strengths and areas for enchancment.

Trying forward, Microsoft AI plans to advance the mannequin additional whereas orchestrating a set of specialised fashions tailor-made to completely different person intents and situations. The division goals to proceed creating main AI options and making them accessible to customers worldwide.

The put up Microsoft Debuts First In-House AI Models: MAI-Voice-1 For Ultra-Fast Speech And MAI-1-Preview For Instruction-Following Tasks appeared first on Metaverse Post.