GLIDE is essential to DALL-E two mainly because it permitted the authors to easily port over GLIDE's text-conditional photorealistic image generation abilities to DALL-E 2 by as an alternative conditioning on image encodings inside the representation space. This point does not originate with DALL-E two (particularly, CLIP demonstrated it https://financefeeds.com/blockdags-community-surges-past-140k-as-50-more-bonus-creates-market-buzz-near-slows-down-polygon-accelerates/