ML//multimodal//text-to-image

Text-to-image is the class of generative models that produce images from text descriptions, of which DALL-E, Midjourney and Stable Diffusion are the best known. It is the technology that made every person a *digital artist* overnight. Its pipeline runs in four stages:

Text-to-image is the class of generative models that produce images from text descriptions, of which DALL-E, Midjourney and Stable Diffusion are the best known. It is the technology that made every person a digital artist overnight. Its pipeline runs in four stages:

1Text encoderusually CLIP2Conditioning3Diffusion4Image decoder