Microsoft Launches MAI-Image-1: First In-House AI Image Generator Debuts in Top 10 Rankings

Microsoft AI's first in-house image generator MAI-Image-1 is now available

Microsoft's first in-house AI image generator, MAI-Image-1, is now available in two products, Bing Image Creator and Copilot Audio Expressions. The company announced the model in October. Microsoft AI chief Mustafa Suleyman wrote in a post on X that the text-to-image model will be "coming soon" to the EU. Suleyman added that the model "really excels at" generating images of food and nature scenes, as well as artsy lighting and photorealistic detail. Microsoft has previously posted more details on its blog: "MAI-Image-1 excels at generating photorealistic imagery, like lighting (e.g., bounce light, reflections), landscapes, and much more. This is particularly so when compared to many larger, slower models. Its combination of speed and quality means users can get their ideas on screen faster, iterate through them quickly, and then transfer their work to other tools to continue refining." Microsoft's MAI-Image-1 will also create AI-generated art to accompany AI-generated audio stories in the "story mode" of Copilot's text-to-speech platform, Copilot Audio Expressions. In August, Microsoft announced their first in-house AI models - the speech model MAI-Voice-1 and the text-based model MAI-1-preview. At the time, the company said it planned to use MAI-1-preview in its Copilot AI assistant in certain unspecified cases, a sign that Microsoft might be pivoting away from its reliance on OpenAI's models. As of today, Microsoft says that its Copilot chatbot is transitioning to OpenAI's latest model GPT-5, while also offering Anthropic's Claude AI models as options to users. MAI-Image-1 is listed as one of the three AI models available on Bing's image creator website and app. The other two models, DALL-E 3 and GPT-4o, are from OpenAI.

Microsoft's new AI image generator already rivals Nano Banana -- here are 5 prompts I used to test it

Microsoft just launched MAI‑Image‑1, the company's first fully in‑house built image generation model. It's already competing at the top of the rankings of major text‑to‑image models like Gemini's Imagen and ChatGPT in the independent site LMArena. For those unfamiliar, LMArena is kind of like the Billboard charts -- but instead of ranking music, it ranks how good AI models are at generating images from text. The public leaderboard is the place where different companies (like Microsoft, Google or open-source projects) submit their text-to-image AI models, and these models are judged based on how well they turn written prompts into high-quality pictures. The judging is based on crowd-voting and pairwise comparisons, which means real people are shown two images generated by different models (without knowing which is which), and they vote on which looks better. The model that consistently wins more matchups climbs the ranks. So when Microsoft says MAI-Image-1 debuted in the Top 10 on LMArena, that means it beat out a lot of competitors in a blind head-to-head test -- and it's now considered one of the best AI image generators available, based on public feedback. Getting started with MAI-Image-1 is easy. You don't need a subscription, just a Microsoft login if you want to keep track of your creations. MAI-Image-1 is also integrated into Copilot in Microsoft Designer and PowerPoint, where you can right-click and use natural language to generate supporting visuals for presentations, social graphics, or concept boards. Microsoft says they prioritized rigorous data selection and nuanced evaluation focused on tasks that closely mirror real‑world creative use cases. Here's the key features mentioned: Use cases for MAI‑Image‑1 are vast considering the tool can be used within Copilot. Try idea prototyping with it by pushing a prompt to create three to five variations until you get what you're looking for, select your favorite, then refine the next prompt based on output. From there, you can export your image and bring into your favorite design tool (Photoshop, Figma, Illustrator) for further editing/branding. Additionally, try incorporating your favorite generated images into blog posts, presentation decks, social posts or mood boards like Mixboard. Here are prompts you can copy/paste into Bing Image Creator or Copilot to test what MAI‑Image‑1 can do. Keep in mind that the image generator will randomly switch from one model to the other, so before you generate, double check to make sure you're actually using MAI‑Image‑1. Prompt: "Create a photorealistic image of a futuristic city skyline at sunset, with reflective glass buildings and flying electric vehicles, in a cinematic wide‑angle style." I have used various versions of this prompt before because it helps me determine how well and image generator handles nature, contrasts and elements that are not yet real (flying vehicles). It tests creativity and nuance. These four images are on-par with top video generators. Prompt: "Generate a close‑up of a gourmet vegan dish plated on a sleek black ceramic plate, soft natural light from the side, minimal background, high resolution." MAI-Image-1 lets you choose how many images you want generated at a time. I wanted to see the different angles and plating of the food, and the images did not disappoint. Prompt: "Create a high contrast marketing hero image for a tech startup: diverse team of four brainstorming around a holographic display, ambient neon lighting, ultra‑wide lens." This image is oddly dark and less clear. Based on this image I would tweak the prompt to encourage more detail and better lighting. Prompt: "Illustrate a serene wildlife scene: a red fox crossing a misty forest clearing at dawn, warm golden hour lighting, ultra‑detailed fur textures, shallow depth of field." This is a stunning image with attention to detail. Looking at this image, it would be hard to determine if this were ChatGPT, Nano Banana or MAI-Image-1. It's easy to see why this image generator is on par with its rivals. Prompt: "Create a visual storyboard: three panels showing the evolution of editing workflow from paper notebooks > laptop screen > holographic AI assistant, seamless transition, clean corporate style." This story board isn't great, but it's a fairly good start. Integrating MAI-Image-1 with something like Mixboard or Nano Banana could help users further their creativity and bring their products and stories to life. MAI‑Image‑1 marks a major step for Microsoft in visual AI. It's accessible, fast and built with real‑world creative workflows in mind. While it's not perfect and the verdict over whether it's as good as Nano Banana is still out, the early results are fairly impressive. You can expect a full MAI-Image-1 vs Nano Banana comparison from me soon. If you're already generating content with AI, it's worth adding MAI‑Image‑1 to your toolkit. Give it a try and let me know what you think in the comments.

Microsoft's new AI image maker is live, and early testers are loving it

What's happened? Back in October, Microsoft AI had unveiled its in-house text-to-image system, named MAI-Image-1. Well, as announced by Microsoft AI CEO Mustafa Suleyman on X, the image-generation model is now live inside Copilot. Touted as a "major improvement" over the previous system, you can start using the model right away -- no long waitlists, no experimental flags. Just open Copilot on the web or desktop, and start generating. MAI-Image-1 is built for photorealistic outputs, tackling historically messy details, especially hands, with much better accuracy. The model supports text-prompt generation and image editing, offering refinement tools similar to what you'd find in DALL-E. Users are already testing it publicly, and early impressions are quite positive. Why this is important: MAI-Image-1 marks a major moment for Microsoft because it's the company's first serious, fully in-house image generator, meaning it no longer leans on third-party models like OpenAI's DALL-E. That means Microsoft can now iterate faster, tailor the tool more closely to Copilot, and better serve regions where competitor models (like Google's Gemini) have limited availability. Early testers are already calling it a "huge improvement", pointing to noticeably better photorealism, more natural lighting, and, most importantly, a major fix to one of AI art's most notorious flaws: weird fingers. Some developers who tried the model say they're impressed with how consistent and responsive it feels, even praising it as "very good" for detailed concept work. Recommended Videos This also signals where Copilot is headed: toward becoming a more complete creative platform rather than just a text assistant. Photorealistic samples shared online, like close-ups of hands in different environments or a realistic coffee cup with visible steam, suggest Microsoft is closing the gap with established AI art tools. That raises the stakes in an increasingly crowded market, where Meta, Google, Grok, and OpenAI are all racing to deliver best-in-class generative visuals. If Microsoft keeps this pace, Copilot could become a go-to space for image creation, not just casual prompting. Why should I care? If you already use Copilot, your image-generation experience just got better automatically, no upgrade required. Just switch to the new model, and you should be good to go. Whether you're creating mood boards, product concepts, thumbnails, illustrations, or just meme-grade chaos, MAI-Image-1 is polished enough to make the process easier and more fun. And because it's integrated directly into Copilot, you don't need to juggle external tools -- everything lives where you're already typing prompts. Okay, so what's next? MAI-Image-1 is already available inside Microsoft's consumer AI tools, such as Copilot and Bing Image Creator -- currently live in most major markets, including the U.S., U.K., Canada, Australia, and other countries where Copilot supports text prompts. However, Mustafa clarified that the model will be "coming soon" to the EU. To try it, simply sign in with your Microsoft account, open Copilot (or Bing Image Creator), and enter a prompt for creating or editing an image. Because the model is built into existing tools rather than a standalone site, it feels like a natural upgrade rather than a separate app to learn. Looking ahead, Microsoft plans to expand the rollout even further, including additional regions, languages, and platform integrations. Early user reactions suggest the model's strength lies in photorealistic detail, refined lighting, and more natural anatomy, so expect Microsoft to refine editing workflows, add more style controls, and possibly open up model choice options. In other words, this launch isn't the finish line -- it's the kickoff of Microsoft staking claim to image generation, and every new iteration could unlock richer creative workflows inside Copilot.