Description

DALL·E is OpenAI’s 12-billion parameter model for creating images from text captions. It is trained on text-image pairs and designed to generate visual content from natural language prompts.

The live page describes DALL·E as a transformer language model that can generate images from scratch and regenerate regions of an existing image in a way that matches the prompt.

How we innovate

DALL·E is distinctive for treating text and images as a single token stream, enabling it to generate images from captions and also extend or revise parts of existing images in a prompt-consistent way.

Use Case / Scenario

1. Text-to-Image Generation

Use DALL·E to turn natural language prompts into images for a wide range of concepts expressible in language.

2. Prompt-Based Image Editing

Use DALL·E to regenerate a rectangular region of an existing image so the edited area stays consistent with the text prompt.

3. Concept Exploration and Visualization

Use DALL·E to explore how text can control visual concepts such as object combinations, perspective, and contextual details.

Visit Website