Microsoft launches its first in-house AI models: MAI-Voice-1 and MAI-1-preview

Microsoft has released two in-house AI models, expanding its AI capabilities for consumer and enterprise contexts. MAI-Voice-1, its first highly expressive and natural speech generation model, is now available in Copilot Daily, Podcasts, and as part of the Copilot Labs experience. This model delivers high-fidelity, expressive audio across both single and multi-speaker scenarios. It stands out for its speed, generating up to one minute of audio in less than a second on a single GPU, making it one of the fastest systems currently offered.

Additionally, Microsoft has introduced MAI-1-preview, its first end-to-end foundation model. It is engineered as a mixture-of-experts architecture and both pre-trained and post-trained on approximately 15,000 NVIDIA H100 GPUs. These technical investments preview enhanced capabilities within Copilot, aiming to handle diverse user instructions and daily informational requests more efficiently.

Looking ahead, Microsoft has indicated its intent to orchestrate a suite of specialized AI models. This strategic focus is intended to unlock broader value by matching a variety of user needs and use cases. The company has signaled strong ambitions for further advancement in artificial intelligence development.