Microsoft unveils AI image editing preview

Microsoft has released a preview version of its AI image generation tool, MAI-Image-2.7, which appears to focus on image editing rather than generating images from scratch. This update marks a shift from the company’s earlier emphasis on photorealism with its initial model, MAI-Image-1.

New model outperforms competitors

According to Neowin, the new version surpasses OpenAI's GPT Image 2.5 Flare and ranks second in the Artificial Analysis evaluation system. This system uses blind human evaluations to compare results without revealing which model produced them, ensuring unbiased assessments.

Editing capabilities and use cases

Microsoft states that the new model allows users to rework existing images by modifying elements such as lighting, object placement, image quality, and text while maintaining consistency and natural integration. The company highlights that the model progresses on 10 use cases and seven types of edits evaluated by Artificial Analysis.

Preview available for testing

The preview version is available for testing on platforms such as Arena, Artificial Analysis, and MAI Playground. For now, only image editing is offered, with text-to-image generation expected to be available soon.