(Image source from: Timesofindia.indiatimes.com)
Microsoft has introduced its first AI image generator called MAI-Image-1. The new model was entirely developed internally by the company based in Redmond. This represents a significant advancement for Microsoft as it aims to lessen its reliance on OpenAI. Microsoft asserts that MAI-Image-1 is designed to benefit creators by avoiding "boring and overly-stylized results." The company claims it gathered input from industry professionals to ensure the AI model produces the most realistic images possible. According to the tech giant, MAI-Image-1 excels at generating "realistic photos," including details like lighting, reflections, and landscapes. The model is also designed to deliver results quickly, enhancing efficiency. Microsoft stated that its AI model is superior in performance compared to “larger and slower models.”
The company has not mentioned any specific model details. Yet, the AI image creation field is primarily led by a few key players, such as OpenAI and Google. Microsoft is working to decrease its reliance on OpenAI, especially after recent changes in their partnership. OpenAI has already released its Sora app on Apple's App Store in the US and Canada, which enables users to create AI-generated videos of themselves that they can share online with the new Sora 2 engine. Meanwhile, Google’s Nano Banana has been gaining attention, with social media users experimenting with various trends thanks to Google's imaging technology, particularly in India, which has the largest user base to date. Microsoft reports that its new image model is already among the top 10 in LMArena, a platform where individuals assess different AI models based on their outputs and vote for their favorites.
MAI-Image-1 joins two other AI models developed by Microsoft, which are the voice generator MAI-Voice-1 AI and the chatbot MAI-1-preview. In addition, the Redmond company has recently begun incorporating AI models from Anthropic into certain features of Microsoft 365. At present, MAI-Image-1 is available for testing on LMArena only. Microsoft has indicated plans to integrate the model into Copilot and Bing Image Creator "very soon."
