Microsoft's Bing Image Creator can now use in-house MAI-Image-1 AI model

Microsoft has rolled out its first in-house AI image generator, MAI-Image-1, to Bing Image Creator and Copilot Audio Expressions

Microsoft has announced that its first in-house AI image generator, MAI-Image-1, is now available in Bing Image Creator and Copilot Audio Expressions. The Windows maker, in a press release, stated that MAI-Image-1 is available on Bing’s website and the Bing mobile app. It can also be accessed from the Bing search bar. The MAI-Image-1 from Microsoft is now an option alongside DALL·E 3 and GPT-4o in the model menu. The image generator model is available in India.

 

Microsoft AI CEO Mustafa Suleyman, in a post on X, said that MAI-Image-1 excels at generating images with artistic lighting/photorealistic detail, nature scenes, and food.

 

For the uninitiated, Bing Image Creator is Microsoft’s AI tool that generates images from text prompts within Bing and Edge, and Copilot Audio Expressions turns written text into expressive audio narration. With MAI-Image-1 integration into Copilot Audio Expressions, when users select Story Mode, Audio Expressions will use MAI-Image-1 to visualise a story with an image.

What is MAI-Image-1, and how was it developed

Microsoft’s MAI-Image-1 is its first in-house text-to-image generation model. According to the company, it has been designed for photorealistic output, and it aims to balance speed and image quality. It is also Microsoft’s third internally developed AI model, following MAI-Voice-1 and MAI-1-preview, which were launched in August to enhance Copilot’s AI capabilities.

 

Microsoft has emphasised MAI-Image-1’s strength in photorealism, highlighting its ability to handle complex lighting effects, such as reflections and bounce light, along with detailed landscape rendering. The company positions the model as faster and more efficient than some of the larger, slower text-to-image systems currently available.

According to Microsoft, MAI-Image-1 was developed with a strong focus on data curation and evaluation for real-world creative applications. The training process incorporated feedback from professionals in the creative sector and included techniques aimed at minimising repetitive or overly simplified artistic patterns in generated images.