Microsoft has begun rolling out its first in-house image generation model, MAI-Image-1, across select platforms including Bing Image Creator and Copilot Audio Expressions. The new tool, which the company says combines speed, realism and creative flexibility, represents a shift towards developing its own artificial intelligence technology rather than relying solely on partners such as OpenAI.
Available now in Bing and Copilot
MAI-Image-1 can now be accessed via Bing Image Creator at bing.com/create, through the Bing mobile app, and directly from the Bing search bar. Users can select MAI-Image-1 alongside other popular image models like DALL-E 3 and GPT-4o, allowing them to compare results and choose the best option for their creative needs.
The model also powers visual storytelling in Copilot Audio Expressions, a feature available through Copilot Labs. In Story Mode, the AI automatically generates unique images to accompany audio stories, combining sound and visuals into one interactive experience.
According to Microsoft, MAI-Image-1 is currently available in all countries that have access to Bing Image Creator and Copilot Labs, with European availability expected “soon”, according to Microsoft AI chief Mustafa Suleyman.
What makes MAI-Image-1 different
Unlike Microsoft’s previous AI image tools, which were built using OpenAI’s DALL-E technology, MAI-Image-1 has been developed entirely in-house by Microsoft AI. The model was trained with feedback from creative professionals to reduce generic results and increase artistic diversity.
Microsoft says MAI-Image-1 excels at creating photorealistic imagery, such as natural lighting effects, landscapes and reflective surfaces, while maintaining a faster rendering time than many larger models. The company highlights that its speed and flexibility make it especially useful for designers and storytellers who want to iterate ideas quickly before refining them in other tools.
Suleyman described the model as particularly strong at generating food and nature scenes, as well as detailed, artistic lighting, areas where photorealism can be challenging for AI systems.
A new direction for Microsoft AI
The release of MAI-Image-1 follows the earlier launch of two other in-house AI models, MAI-Voice-1, which generates expressive and natural speech, and MAI-1-preview, a large language model for text-based tasks. Together, these models mark Microsoft’s growing independence in AI research and development, building on its partnership with OpenAI while expanding its own capabilities.
Microsoft says the MAI family of models supports its broader mission to create “AI for everyone”, technology designed to be supportive, helpful and responsible. The company’s approach combines multiple specialised models to meet different user needs, from speech and storytelling to image creation and beyond.
When to expect wider rollout
While MAI-Image-1 is currently available in Bing and Copilot Labs, Microsoft plans to bring it to more of its products over the coming months. Future updates to the Copilot assistant across Microsoft 365 apps are expected to include MAI models, alongside continued support for OpenAI and Anthropic systems.
For now, users can test the new model for free through Bing Image Creator or Copilot Audio Expressions, experimenting with prompts to see how MAI-Image-1 compares to DALL-E 3 and other AI image tools.
At a time when artificial intelligence continues to shape how people create and communicate, Microsoft’s latest step signals a commitment to developing its own foundation models, ones that are not just faster and more capable but also designed to reflect human creativity more closely.
