Microsoft AI announced that it has built its first text-to-image generator designed and developed in-house. The company will integrate the tool into its suite of generative artificial intelligence products.

While Microsoft has offered image generator tools in the past for products like Copilot and Bing Image Creator, these were previously powered by models from OpenAI.

"MAI-Image-1" marks a significant step toward developing its own proprietary AI models based on its technology. It can provide photorealistic imagery such as lightning and landscapes, and is the next step in its journey, the company wrote in a blog post. 

The model can process requests and produce images faster than “larger, slower models,” and has secured a spot in the top 10 in LMArena, a community platform and AI benchmark site where people can compare the output from different systems and vote on which they determine to be the best.

The launch of MAI-Image-1 comes as executives at the company rethink their relationship with OpenAI, a startup Microsoft funded.

Microsoft recently integrated Anthropic’s AI models into some Microsoft 365 features and used its large language models to train its own model, MAI-Image-1.

MAI-Image-1 was developed with input from creative professionals, specifically to overcome the common pitfall of "repetitive or generically stylized outputs."

Microsoft trained the model with the goal to support creators.

For example, the company's engineers prioritized data selection and nuanced evaluation focused on tasks that closely mirror real-world creative use cases.

Engineers took into account feedback from professionals in the creative industries. This model is designed to deliver flexibility, visual diversity and practical value.