Microsoft Unveils Three In-House AI Models Targeting Speech, Voice, and Image Generation
Microsoft has introduced MAI-Transcribe-1, MAI-Voice-1, and MAI-Image-2, its first internally developed AI models focusing on transcription, voice generation, and image creation, respectively.
Microsoft has announced the launch of three new AI models developed entirely in-house: MAI-Transcribe-1, MAI-Voice-1, and MAI-Image-2. These models are designed to enhance capabilities in speech transcription, voice generation, and image creation, respectively.
MAI-Transcribe-1 is a speech-to-text model that achieves a 3.8% Word Error Rate across 25 languages, outperforming existing models from OpenAI and Google. It supports multiple audio formats and offers faster transcription speeds compared to current Microsoft Azure offerings.
MAI-Voice-1 is a text-to-speech model capable of generating 60 seconds of natural-sounding audio in a single second. It preserves speaker identity across long-form content and supports custom voice creation from brief audio samples. The model is priced at $22 per 1 million characters.
MAI-Image-2, the upgraded image creator, delivers faster generation times and has been integrated into Bing and PowerPoint. It is priced at $5 per 1 million tokens for text input and $33 per 1 million tokens for image output.
These models are available through Microsoft Foundry and the new MAI Playground, marking a significant step in Microsoft's AI development efforts.
