MAI's 7 New Models

Microsoft has released a new set of AI tools called MAI-Transcribe-1, MAI-Voice-1, and MAI-Image-2. These models are part of Microsoft Foundry and are designed to help developers build better applications. They are also used inside Microsoft products like Copilot and Bing to make them smarter and faster.

Benefits

These new models offer significant advantages for businesses and creators. MAI-Transcribe-1 provides very accurate speech recognition across 25 different languages. It costs about half as much to run as other popular transcription tools. MAI-Voice-1 creates high-quality audio extremely quickly. It can generate a full minute of expressive speech in less than one second using just one computer chip. MAI-Image-2 is a powerful tool for creating images from text descriptions. It ranked third on a major leaderboard for image generation models. Together, these tools allow companies to build voice assistants, create custom voices, and generate marketing images without needing expensive hardware.

Use Cases

Developers can use these models in many real-world situations. For customer service, companies can use the transcription model to automatically summarize phone calls and understand what customers are saying. In education, schools can use these tools to create captions for lectures and turn spoken lessons into written notes. Content creators can use MAI-Voice-1 to generate realistic voiceovers for videos or podcasts. Marketing teams can use MAI-Image-2 to design new logos, social media posts, or product mockups instantly. The tools are also useful for making events more accessible by providing real-time captions for large gatherings or meetings.

Pricing

Pricing details for these models are not available in the provided information. The models are available exclusively on Microsoft Foundry for developers to use. Custom voice creation features may require an approval process and specific Azure Speech subscriptions.

Vibes

Users and partners have responded very positively to these new tools. Rob Reilly, the Global Chief Creative Officer at WPP, called MAI-Image-2 a genuine game-changer. He noted that the model respects the craft involved in creating real-world images for campaigns. Other developers are excited about the speed and accuracy, especially the ability to generate a minute of audio in under a second. The community sees these tools as a major step forward for building enterprise-grade AI applications.

Additional Information

Microsoft developed these models in close collaboration with photographers, designers, and visual storytellers. This partnership ensured the tools meet professional standards. MAI-Transcribe-1 and MAI-Voice-1 are designed to work together to create end-to-end voice experiences. The models are optimized for efficiency and security. They are currently in public preview and are available for developers to experiment with through the MAI Playground or to deploy in Microsoft Foundry.