Images from a description
Image generation AI produces illustrations and photorealistic images from an instruction such as a cat walking on a beach at sunset, watercolor style. It is among the most striking areas of generative AI: high-quality images in seconds, with no design training required.
The main services
- Midjourney: regarded for artistic quality
- Stable Diffusion: open and free, and able to run on your own machine
- Image generation in ChatGPT and Gemini: convenient alongside a conversation
- AI built into design tools such as Canva: integrated with producing documents and social images
Starting with the image generation inside a chat assistant you already use is the easiest route.
How the images get made
Most of these run on diffusion models. Starting from a noise image, they remove noise gradually in a direction that matches the instruction until a picture emerges. Training on enormous numbers of image and caption pairs is what teaches the correspondence between words and pictures.
Getting better results
Separate what you want into subject, setting, style, and mood. Not just a cat, but a calico cat napping by a window, soft morning light, picture-book style. Layering description gets you closer. Rather than settling for the first result, point out what is wrong and regenerate — that is the normal way to work.
Many services also support negative prompts for what you want excluded. Saying no text or no people up front helps stabilize the result.
Rights and cautions
Rights in generated images differ by service, and the legal position is still being argued. Avoid images resembling real people or existing characters, which invite trouble. For professional use, read the terms of the service you are using once, carefully.