Microsoft Unveils Three AI Models to Enhance Speech and Image Processing

Microsoft has announced the release of three proprietary AI models: MAI-Transcribe-1, MAI-Voice-1, and MAI-Image-2. These models are now available through Microsoft Foundry and the MAI Playground.

MAI-Transcribe-1 is designed for speech-to-text tasks, supporting transcription across 25 widely spoken languages. It offers batch transcription speeds 2.5 times faster than Microsoft's existing Azure Fast service. MAI-Voice-1 focuses on voice generation, producing natural-sounding speech with emotional nuance and consistency over extended content. It also allows for the creation of custom voices from short audio samples. MAI-Image-2 is a text-to-image generation model that delivers realistic lighting, accurate skin tones, and clear text rendering within images. It has been integrated into services such as Bing and PowerPoint.

These models are available for commercial use, with pricing set below comparable offerings from competitors. MAI-Transcribe-1 starts at $0.36 per hour, MAI-Voice-1 at $22 per one million characters, and MAI-Image-2 at $5 per one million tokens for text input and $33 per one million tokens for image output.

Despite developing these in-house models, Microsoft remains committed to its partnership with OpenAI. Mustafa Suleyman, CEO of Microsoft AI, emphasized the company's dedication to this collaboration while pursuing parallel AI development strategies.