# DALL-E 2

DALL-E 2 is a text-to-image model developed by OpenAI, released in 2022, that generates realistic images from natural language prompts using diffusion and CLIP embeddings.

DALL-E 2 is a text-to-image model developed by [OpenAI](https://www.wikiprompt.org/wiki/openai) and released in 2022. It generates digital images from natural language descriptions, known as prompts, using [deep learning](https://www.wikiprompt.org/wiki/deep-learning) methodologies. The model is a successor to the original DALL-E and was designed to produce more realistic images at higher resolutions, with the ability to combine concepts, attributes, and styles.

DALL-E 2 is part of the broader field of [generative artificial intelligence](https://www.wikiprompt.org/wiki/generative-ai), which focuses on creating new content. It uses a diffusion-based architecture, a departure from the autoregressive approach of its predecessor. The name is a portmanteau of the Pixar robot WALL-E and the Spanish surrealist artist Salvador Dalí.

## History and background

The original DALL-E was announced by OpenAI in a blog post on 5 January 2021, using a modified version of [GPT-3](https://www.wikiprompt.org/wiki/transformer) to generate images. On 6 April 2022, OpenAI announced DALL-E 2 as a successor, emphasizing its ability to generate more realistic imagery and combine concepts. Initially, access was restricted to pre-selected users for a research preview due to ethical and safety concerns. On 20 July 2022, DALL-E 2 entered a beta phase, inviting 1 million waitlisted individuals; users could generate a limited number of images for free each month and purchase additional credits. The waitlist requirement was removed on 28 September 2022, opening the model to the general public.

In early November 2022, OpenAI released DALL-E 2 as an API, allowing developers to integrate the model into applications. The API operates on a cost-per-image basis, with prices varying by resolution and volume discounts for enterprise clients. Microsoft integrated DALL-E 2 into its Designer app and Image Creator tool in Bing and Microsoft Edge. In February 2024, OpenAI began adding watermarks to DALL-E generated images, embedding metadata using the C2PA (Coalition for Content Provenance and Authenticity) standard.

## Technology

DALL-E 2 uses 3.5 billion parameters, fewer than its predecessor's 12 billion. Instead of an autoregressive transformer, it employs a diffusion model conditioned on CLIP image embeddings. During inference, a prior model generates these image embeddings from CLIP text embeddings. This architecture is similar to that of [Stable Diffusion](https://www.wikiprompt.org/wiki/stable-diffusion), released a few months later. The model leverages [neural networks](https://www.wikiprompt.org/wiki/neural-network) and [machine learning](https://www.wikiprompt.org/wiki/machine-learning) techniques, building on advancements in [artificial intelligence](https://www.wikiprompt.org/wiki/artificial-intelligence) research.

CLIP (Contrastive Language-Image Pre-training) is a separate model trained on 400 million image-text pairs from the internet. It plays a crucial role in understanding and ranking DALL-E 2's outputs, selecting the image that best matches the prompt. The diffusion process iteratively refines random noise into a coherent image, guided by the CLIP embeddings.

## Capabilities

DALL-E 2 can generate imagery in multiple styles, including photorealistic images, paintings, and emoji. It can manipulate and rearrange objects, and correctly place design elements in novel compositions without explicit instruction. For example, when asked to draw a daikon radish blowing its nose, sipping a latte, or riding a unicycle, the model often places handkerchiefs, hands, and feet in plausible locations. It can also infer appropriate details, such as adding Christmas imagery to prompts associated with the holiday or rendering shadows that were not mentioned.

The model exhibits broad understanding of visual and design trends, and can produce images for a wide variety of arbitrary descriptions from various viewpoints, with only rare failures. Its visual reasoning ability is sufficient to solve Raven's Matrices, a test often used to measure human intelligence.

### Image modification

DALL-E 2 can produce variations of existing images, as well as edit images to modify or expand them. The inpainting and outpainting abilities use context from the original image to fill in missing areas in a consistent medium, following a given prompt. For instance, users can insert new subjects into an image or extend it beyond its original borders. According to OpenAI, outpainting considers existing visual elements such as shadows, reflections, and textures.

## Impact and legacy

DALL-E 2 influenced subsequent developments in text-to-image generation, including DALL-E 3, released in October 2023, which was integrated into ChatGPT and later replaced by GPT Image in March 2025. The model also contributed to broader discussions about the capabilities and ethical implications of generative AI, including concerns about misuse and the need for provenance measures like watermarking.

---
Source: https://www.wikiprompt.org/wiki/dall-e-2
License: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/)
Last updated: 2026-09-12T16:28:33.396184+00:00