DALL-E
By OpenAI
DALL-E is a series of text-to-image generation models developed by OpenAI that create original images from natural-language text descriptions, and is integrated directly into ChatGPT for conversational image generation and editing.
Definition
DALL-E is a series of text-to-image generation models developed by OpenAI that create original images from natural-language text descriptions, and is integrated directly into ChatGPT for conversational image generation and editing.
Overview
OpenAI introduced the original DALL-E in January 2021, demonstrating that a model trained to predict text and image tokens together could generate plausible, often surprising images from short prompts. DALL-E 2 followed in 2022 with sharper, more photorealistic output and inpainting/editing capabilities, and DALL-E 3 arrived in 2023 with much stronger prompt adherence, largely by pairing the image model with more detailed, model-generated captions during training. Like other diffusion model-based systems such as Stable Diffusion and Midjourney, DALL-E turns a text prompt into a numerical representation and progressively refines random noise into a coherent image guided by that representation. DALL-E 3's tight integration with ChatGPT is a key differentiator — users can iterate on an image through natural conversation, asking ChatGPT to adjust colors, composition, or style without needing precise prompt-engineering syntax. DALL-E popularized text-to-image generation for a mainstream audience and helped establish generative AI as a category, alongside competitors that have since matured their own strengths in realism, artistic style, and controllability.
Key Features
- Text-to-image generation from natural-language prompts
- Deep integration with ChatGPT for conversational, iterative image editing
- Inpainting and outpainting to edit or extend parts of an image
- Strong prompt adherence in DALL-E 3 via detailed auto-generated captions during training
- API access for developers to embed image generation in applications
- Built-in safety filters restricting certain categories of generated content
- Support for varied artistic styles, from photorealism to illustration