![]() |
Text generators are trained on large amounts of text from books, articles, and websites which is analyzed to find patterns and relationships and create new texts by predicting the word or sentence most likely to follow another in a sequence. Text generators can be used to produce a wide variety of content including essays, memos, brochures, poems, songs, and screenplays.
Examples: ChatGPT | Copilot | Gemeni |
![]() |
Image generators learn by analyzing sets of images with captions or text descriptions. Once they learn which images are associated with which concepts, they can combine them to create new images in a range of styles from photorealistic to abstract.
Examples: Dall-E 2 | Midjourney | Stable Diffusion |
Source: Dall-E 2
Prompt: Paint a portrait of Homer Simpson in the style of Edvard Munch's "The Scream"
![]() |
Video generators learn by analyzing large sets of annotated video and generating new video in response to a text prompt. Alternatively, users can upload existing videos and edit them using text prompts or by applying canned filters and effects.
Examples: Synthesia | Pictory | Kapwing |
Source: Kapwing
Prompt: Create a video on how to make a sandwich
![]() |
Code generators use algorithms trained on existing source code—typically produced by open source projects for public use—and generate new code based on those examples. Some tools can also analyze and debug existing code or offer suggestions for improvement. Examples: CodePal | Tabnine | GitHub Copilot |