Microsoft has introduced its first in-house image generation model, MAI-Image-1. The model currently ranks ninth among text-to-image systems on LMArena. According to Microsoft AI, MAI-Image-1 was designed to avoid repetitive or overly generic results. The team incorporated feedback from creative industry professionals to fine-tune the model. It performs especially well at producing photorealistic images, capturing details like lighting effects and landscapes, and runs faster than many larger models.
The model is available for testing on LMArena and will soon be integrated into Copilot and Bing Image Creator. Microsoft emphasizes its focus on safe and responsible outputs. MAI-Image-1 expands the company’s own AI lineup, which also includes the speech generation model MAI-Voice-1 and the chatbot MAI-1-preview.
