dall-e-2
by openai
Pricing
Input
$0.00 / 1M tokens
Output
$0.01 / 1M tokens
DALL·E 2 Model Suite by OpenAI
Overview
DALL·E 2 is a powerful text-to-image model developed by OpenAI, released in April 2022. Based on a diffusion architecture conditioned on CLIP embeddings, DALL·E 2 can generate highly detailed, high-resolution (1024×1024) images from text prompts. It also supports advanced image editing features such as inpainting and outpainting.
Features
- Generates photorealistic and abstract images
- Inpainting and outpainting for creative editing
- High resolution (1024×1024), 4× sharper than original DALL·E
- Guided by CLIP for better text-image alignment
Use Cases
- Marketing and social media content creation
- Storyboarding and concept art for design
- Educational illustrations and visual aids
- AI-assisted photo editing (e.g. object replacement)
Limitations
- Struggles with object relationships and prompt negations
- Generated text within images is often unreadable
- Biases in output (e.g. gender, race stereotypes)
- Faces restricted for privacy and ethics compliance
Performance Comparison
| Model | Resolution | Editing Features | Prompt Fidelity | Image Realism |
|---|---|---|---|---|
| DALL·E 1 | 256×256 | None | Basic | Moderate |
| DALL·E 2 | 1024×1024 | Inpainting, Outpainting | Improved | High |
| DALL·E 3 | 1024×1024+ | + Context-aware editing | Very High | Very High |