Skip to main content

Image Generation

Welcome to part 2 of this workshop, where we will interact with a text-to-image model, DALL-E 3. Start by clicking on the Images playground, and selecting dall-e-3 from the deployments drop-down menu.

Creating our first images​

What is DALL-E 3?

DALL-E 3 is a neural network based model that can generate graphical data from natural language input. Put more simply, you can provide DALL-E 3 with a description and it can generate an appropriate image.

Let's begin with generating an image by entering a basic prompt in the description box, then clicking on generate:

Enter in the user prompt:
A corgi practicing karate

Images generated by DALL-E 3 are original; they are not retrieved from a curated image catalog. In other words, DALL-E 3 is not a search system for finding appropriate images - it is an artificial intelligence (AI) model that generates new images based on the data on which it was trained.

Be specific​

Details allow for more accurate responses. Similar to text generation, image generation models greatly benefit from detailed descriptions of what you are looking to generate.

For example, enter the following prompt:

Enter in the user prompt:
An elephant on a skateboard

Then select Generate and view the image that is generated.

Now, let's modify the prompt by adding more details to the description:

Enter in the user prompt:
A purple elephant on a skateboard performing an ollie, in the style of Picasso

Select Generate once again and compare the results.

Best practices​

To create effective and accurate images with DALL-E 3, here are some best practices to follow:

  1. Clear and descriptive prompts: Craft your text prompts to be clear and detailed. The more specific you are with your description, the more likely DALL-E 3 will generate an image that matches your request. Include attributes such as the subject, action, environment, style, and any important details.

  2. Use of adjectives: Employ adjectives and adverbs to describe the qualities, emotions, and characteristics you want the image to convey. This helps in refining the generated image to better match your vision.

  3. Balance detail with simplicity: While details are important, overly complicated or contradictory prompts can confuse the AI, leading to unexpected results. Aim for a balance where your description provides enough context without being overly convoluted.

  4. Experiment with different styles: Specify artistic styles or influences if you want your image to have a particular aesthetic. For example, you could ask for an image in the style of Van Gogh or futuristic concept art.

  5. Iterative approach: Often, the first image generated may not be perfect. Use it as a starting point and iteratively refine your prompt based on the output to get closer to your desired result.

  6. Aspect ratio and composition: If you have a preference for the image’s composition or aspect ratio, include it in your prompt. For example, you might request a landscape-oriented image or a portrait with a subject off-center.

  7. Cultural and contextual references: If appropriate, include cultural or historical references to provide additional context that can help guide the image generation process.

  8. Ethical considerations: Be mindful of the ethical implications of your prompts. Avoid creating images that are offensive, perpetuate stereotypes, or infringe on copyrights.

  9. Testing and learning: Experiment with different prompts to understand how DALL-E 3 interprets various descriptions. This learning process can help you improve the precision of your prompts over time.

  10. Following guidelines: Adhere to OpenAI's use-case policy and content guidelines when creating prompts. Avoid requesting images that are not allowed as per OpenAI's content policy.