Skip to main content
DALL-E 2 is OpenAI’s advanced text-to-image foundation model, designed to generate high-quality images from both textual and visual prompts. Unlike large language models such as GPT-3, DALL-E 2 specializes in creative image synthesis. You can:
  1. Provide a text prompt describing the scene, style, or concept you want to visualize.
  2. Supply an image prompt to create variations, overlays, or edits of an existing picture.
The model processes your input and returns one or more fully generated images, offering endless possibilities for design, prototyping, and creative exploration.
The image is a diagram illustrating the DALL-E 2 Foundation Model, showing input as text and images leading to the model, which then outputs images.

​
Key DALL-E 2 APIs

DALL-E 2 provides three core RESTful endpoints for image creation and manipulation:
APIDescriptionTypical Use Case
Image GenerationGenerate new images from scratch based on a descriptive text prompt.Concept art, storyboarding, product mockups
Image EditingEdit or extend an existing image by applying a mask and text prompt to specify changes.Photo retouching, add/remove objects
Image VariationProduce multiple stylistic variations of a source image without any additional text input.Branding explorations, style testing
The image outlines three key APIs: Image Generation, Image Editing, and Image Variation, with a brief description of the Image Editing API's functionality.

​
1. Image Generation

Create brand-new visuals by sending a textual description: 
from openai import OpenAI

client = OpenAI()

response = client.images.generate(
    prompt="A futuristic city skyline at sunset, neon-lit skyscrapers reflecting on water",
    n=1,
    size="1024x1024"
)

print(response.data[0].url)

​
2. Image Editing

Modify a specific region of an existing image by supplying:
  • image: the original image file
  • mask: a black-and-white mask highlighting the edit area
  • prompt: text describing what to place in the masked region
response = client.images.edit(
    image=open("original.png", "rb"),
    mask=open("mask.png", "rb"),
    prompt="Add a sleek flying car in the sky above the buildings",
    n=1,
    size="512x512"
)

print(response.data[0].url)
Ensure your mask file aligns exactly with the dimensions of the source image. Mismatched sizes will result in API errors.

​
3. Image Variation

Generate several stylistic renditions of an existing image: 
response = client.images.variations(
    image=open("input.png", "rb"),
    n=3,
    size="256x256"
)

for img in response.data:
    print(img.url)
DALL-E 2 supports three output resolutions: 256×256, 512×512, and 1024×1024. Square images tend to produce the best results, but feel free to experiment with other aspect ratios.
The image describes three key APIs: Image Generation, Image Editing, and Image Variation, with a note that Image Variation generates multiple variations of the same image in different sizes.
Remember, generative image models incorporate a degree of creative randomness. Iterating your prompts—tweaking style descriptors, color palettes, or composition details—will help you achieve the perfect result.

​
Next Steps: Live Demo

In the following section, we’ll walk through a hands-on demonstration of each DALL-E 2 API, complete with live code execution and real-time image generation.

Watch Video