Skip to main content
Learn how to generate stunning images from text prompts using the OpenAI DALL·E API. You can customize the prompt, choose the number of outputs, and select the resolution that best fits your application.

​
Table of Contents

  1. Prerequisites
  2. Setup and Helper Function
  3. Supported Image Sizes
  4. Generate and Display a Single Image
  5. Generate Multiple Images
  6. Base64-Encoded Output
  7. Integrating with GPT Models
  8. References

​
Prerequisites

  • Python 3.7+
  • An OpenAI API key
  • openai Python package
Install the SDK with: 
pip install openai
Store your API key in an environment variable for security:
export OPENAI_API_KEY="your_api_key_here"

​
Setup and Helper Function

Import packages, configure your key, and wrap the DALL·E call in a reusable function: 
import os
import openai
from IPython.display import Image, display

# Load API key from environment
openai.api_key = os.getenv("OPENAI_API_KEY")
# Alternatively, set directly (not recommended for production)
def generate_image(prompt: str, size: str = "512x512", n: int = 1, response_format: str = "url"):
    """
    Generate `n` images for a given text prompt.

    Args:
      prompt: Text description to guide image creation.
      size:  Resolution, choose from our supported list.
      n:      Number of images to generate.
      response_format: "url" or "b64_json".

    Returns:
      A list of dicts containing the image data (URL or base64 JSON).
    """
    response = openai.Image.create(
        prompt=prompt,
        n=n,
        size=size,
        response_format=response_format
    )
    return response["data"]

​
Supported Image Sizes

Choose the resolution that matches your use case:
SizePixel DimensionsUse Case
Small256×256Thumbnails, icons
Medium512×512Social posts, blog headers
High1024×1024Print, high-resolution display

​
Generate and Display a Single Image

Create a 512×512 image of a “cozy coffee-shop corner” and render it in Jupyter: 
# Generate one image
images = generate_image(
    prompt="Interior photography of a cozy corner in a coffee shop, using natural light.",
    size="512x512",
    n=1
)

# Display the first result
image_url = images[0]["url"]
Image(url=image_url, width=512)
Each execution produces a new, unique image.

​
Generate Multiple Images

Request several variations at once by increasing n and iterating over the response: 
# Generate three variations
images = generate_image(
    prompt="Interior photography of a cozy corner in a coffee shop, using natural light.",
    size="512x512",
    n=3
)

# Display all generated images side by side
for img in images:
    display(Image(url=img["url"], width=256))

​
Base64-Encoded Output

If you prefer embedding images directly (e.g., in HTML or JSON), request b64_json: 
# Generate one base64 image
b64_data = generate_image(
    prompt="A futuristic city skyline at dusk, cinematic lighting",
    size="1024x1024",
    n=1,
    response_format="b64_json"
)

# Access the base64 string
encoded_str = b64_data[0]["b64_json"]
Large base64 payloads can impact performance. Use b64_json sparingly.

​
Integrating with GPT Models

You can automate end-to-end content creation by combining DALL·E with GPT:
  1. Text-to-Image Workflow
    • Use GPT-3.5 Turbo to draft descriptive prompts.
    • Feed them into DALL·E for image generation.
  2. Prompt Engineering
    • Send long-form descriptions to GPT-4 to refine or expand for DALL·E.
    • Obtain optimized prompts that yield better visual results.
This approach can power dynamic blog posts, marketing assets, or social media content.

​
References

Watch Video

Practice Lab