Type a sentence and get an image. Feed a prompt and get a poem, a song, or a block of code. Generative AI covers systems that create new content rather than merely classifying or predicting. The output is novel, though it derives from patterns in training data.

Large language models generate text. Diffusion models generate images. Audio models generate speech and music. Video models are improving quickly. Each uses different architectures, but all learn distributions of data and sample from them.

Common generative outputs

• Text, from articles to code
• Images, from photos to illustrations
• Audio, including speech and music
• Video and animation
• 3D models and design files

The field raises difficult questions. Copyright, attribution, and misinformation all complicate deployment. Generated content can be convincing and false at the same time, which puts pressure on platforms and regulators to respond.