Definition
Plain language
Typing a description and getting a picture back.
As stated in the literature
Generative image synthesis conditioned on a text prompt; struggles with fine-grained specificity such as exact species or entity identity, which is why generated fakes often fail to be retrieved for the intended query.
Also called: text-to-image generator, text to image
Why it matters: It makes convincing images cheap to produce, but its weakness at pinning down exact identities means fully generated fakes often don't match what a search was actually looking for.
For example, you type "a cat wearing a tiny wizard hat" and get back a picture that never existed.
Heard on the show
“You take a text-to-image generator, you type "a blue-chinned sapphire hummingbird with a thick beak," you get a picture, you upload it to the image library.”Episode 247 — One Edited Photo, an Honest Caption, and a RAG System That Believes It