DALL·E 2

DALL·E 2 was OpenAI's second text-to-image system, announced in April 2022, a year after the original DALL·E. The research paper behind it, published on 13 April 2022, describes the architecture the company called unCLIP: one network turns the written prompt into a CLIP image embedding, a second one — a diffusion decoder — turns that embedding into the picture. OpenAI's own comparison put DALL·E 2 ahead of its predecessor in 71.7 percent of judgements on how well the image matched the caption and in 88.8 percent on photorealism, at four times the resolution. Its real significance is what happened around it rather than inside it. OpenAI let users in gradually: trusted testers first, a beta from July 2022, then no waiting list at all. By the time the API opened in public beta in November 2022, the company said more than three million people were already using DALL·E and generating over four million images a day — numbers no image generator had reached before. Microsoft built it into the Designer app; that is the line that runs from here to today's image tools inside office software. It was also the first widely available model that could edit an existing image rather than only produce a new one: DALL·E 2 supported inpainting, painting over a masked region of an uploaded picture and filling it in from a prompt. Every 'edit this photo' feature in later systems descends from that endpoint. OpenAI announced the deprecation of the DALL·E snapshots on 14 November 2025 and removed both dall-e-2 and dall-e-3 from the API on 12 May 2026, pointing developers to the GPT Image family. The model is kept here as a historical entry; current capability lives in the GPT Image 2 profile.

▸Videos

No videos yet.

▸Reviews