System Design of a Text-to-Image Product at Scale

Know the Vocabulary

This lesson uses the terms below many times. Read them once now, and come back here when a later section needs one.

• Text-to-image product: A product where a user types a short text, called a prompt, and gets back images that match it.
• Diffusion model: The kind of model most image products use. It starts from random visual noise and cleans it up over many denoising steps, until a picture appears.
• Denoising step: One pass of the diffusion model over the picture, which removes some of the noise