AI image generators like DALLE-2 and Microsoft Designer use machine learning algorithms to create images from natural language descriptions. These models are trained on large datasets of text-image pairs, allowing them to learn the relationship between language and images. When a user inputs a text prompt, the AI system uses its knowledge of this relationship to generate a corresponding image. The generated image is created using a process called stable diffusion, which involves transforming the text prompt into a series of numbers that represent the image. These numbers are then used to create a digital image that is displayed to the user. The quality and accuracy of the generated image depend on the size and diversity of the training dataset, as well as the complexity of the AI model.