DALL-E 2 is the second-generation image generation model from OpenAI, released in April 2022 as a significant upgrade over the original DALL-E. It produces higher-resolution images with better coherence, more accurate prompt adherence, and the ability to perform advanced tasks such as inpainting, outpainting, and image-to-image transformations.

The model introduced a number of technical improvements, including a diffusion-based architecture that replaced the original autoregressive approach, resulting in sharper images with fewer artifacts and more consistent handling of complex prompts. DALL-E 2 could generate images at 1024x1024 resolution, a substantial increase over its predecessor, and showed better understanding of spatial relationships, object attributes, and compositional structure. The model also included safety features designed to prevent the generation of harmful, inappropriate, or infringing content.

DALL-E 2's release brought text-to-image generation into mainstream creative workflows, with the model being adopted by designers, marketers, illustrators, and hobbyists for everything from concept exploration to final production assets. While newer models have since surpassed it in various capabilities, DALL-E 2 remains historically significant as one of the models that established text-to-image generation as a practical creative tool rather than a research novelty.