Screenshot by Lance Whitney/ZDNET
Follow ZDNET: Add us as a preferred source on Google.
ZDNET's key takeaways
Microsoft is now testing its own in-house AI image generation model.
You can test the new model at the LMArena leaderboard site.
The new model will soon be available in Copilot and Bing Image Creator.
Ever use Microsoft Copilot or Bing Image Creator to generate AI-based images? The models that Microsoft currently uses for this task are actually from OpenAI, specifically GPT-4o and DALL-E-3. But now Microsoft has cooked up its own image generator model as it shifts more of its AI development in house.
In a blog post published Monday, Microsoft introduced its new image generator model dubbed MAI-Image-1. In training this new model, the company boasted that it strived to avoid repetitive or generic output. To assist with this goal, Microsoft solicited feedback from creative professionals. The end result should be images that more closely mirror real-world examples.
"MAI-Image-1 excels at generating photorealistic imagery, like lighting (e.g., bounce light, reflections), landscapes, and much more," Microsoft said in its post. "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."
The new model isn't accessible through Copilot or Bing Image Creator, at least not yet. But it has made its first mark by scoring in the top ten at AI leaderboard LMArena. Here, people can test and rate different AI models to grade their effectiveness. The more votes a particular model gets, the higher it appears in the rankings. As of now, MAI-Image-1 is No. 9 on the board, not a bad start for a brand new model.
MAI-Image-1 is just Microsoft's latest in-house AI model now being tested. In August, the company released MAI-Voice-1 as a natural speech generation model and MAI-1-preview as a large-language model for general text generation. Like MAI-Image-1, both of these models are available for testing at LMArena.
"We are committed to ensuring safe and responsible outcomes," Microsoft said about MAI-Image-1. "That has driven us to begin testing this model in LMArena so that we can gather insights and feedback. We're excited to be making MAI-Image-1 available in Copilot and Bing Image Creator very soon. For now, give it a try in LMArena and let us know what you think."
To give MAI-Image-1 a whirl, browse to the image generator page at LMArena. Click the drop-down menu at the top and change the selection from Battle (which chooses the models for you) to Direct Chat. Then click the menu to select the model and set it to mai-image-1. At the prompt, type a description of the type of image you want to generate and check out what the model creates.
Screenshot by Lance Whitney/ZDNET
Next, you might want to see how MAI-Image-1 compares with other image generator models. For that, click the first drop-down menu and change the selection from Direct Chat to Side by Side. Choose MAI-Image-1 as the first model if it's not already chosen. Then pick a different model for the competitor, maybe DALL-E-3 as that's the model Microsoft currently uses. Again, type a description at the prompt and see which image you prefer.
Screenshot by Lance Whitney/ZDNET
The relationship between Microsoft and OpenAI had been a cozy one in the past. Microsoft kicked in much of the funds to help OpenAI develop its AIs and win a name for itself in the industry. On the flip side, OpenAI provided Microsoft with the underlying models and technology to help it integrate AI into its core products and services.
Recently, though, the two companies have drifted apart. OpenAI is increasingly working with other partners, while Microsoft is bringing more of its AI expansion in house.