Just as with text-based AI tools, image generation tools such as Adobe Firefly rely on a user’s prompts to create good images. After you enter a prompt, the AI model analyzes your input and generates a response based on the patterns it has learned through its training. More descriptive prompts can improve the quality of the outputs. This guide gives you a basic overview of how to generate prompts for image-based AI tools.
-
As with text-based tools, before you start using an image-based AI tool for your Harvard work, you should be sure to review the University’s guidelines. Keep in mind that AI image tools can be extremely variable in quality and can produce offensive content, so be sure to carefully review any work containing AI content before you use or publish it.
-
Pay particular attention to copyright issues surrounding AI-generated images. Firefly is trained on Adobe Stock images, openly licensed content, and public domain content where copyright has expired, making it safe for use without infringing on copyright. For this reason, we recommend Harvard faculty, students, and staff use Firefly for experimentation with AI images.
- Consider AI’s limitations: Just like text-based AI tools can “hallucinate,” image-based tools can add strange artifacts to your images, particularly when dealing with people or places. Look for things like bizarre expressions on the faces of people and animals, contorted bodies, additional fingers on hands, or features that look out of place in a particular location such as palm trees and sandy beaches in Harvard Yard.
Getting started with image prompts
A basic formula of “subject + style + details + format of output” will get you started creating AI images. Keep adding more detail to get better results.
Describe the subject in as much detail as you can
Asking AI to “draw a cat” is so broad that you’ll either get a very generic image or a wide variety of approaches:
To get something closer to what you want, ask yourself questions such as: What is the cat doing? What color is it? What mood is it in? Where is it and what is in the background? By adding these to your prompt you’ll get much closer to the image you’re imagining, e.g. “Draw a ginger-and-white striped cat looking excited as it chases a mouse around a kitchen”.
Specify the style of image
Generative AI can create images in the style of photographs, paintings, cartoons, illustrations, pencil drawings, and much more. You can even try specific artistic techniques such as pointillism, pop art, cubism, impressionism, etc. Try adding a style to your prompt, i.e. “Draw a ginger-and-white striped cat looking excited as it chases a mouse around a kitchen, in the style of an impressionist painter.”
Add more details and refine
Other considerations like lighting, the positioning of the subject within the frame, background details, primary color scheme, degree of realism, and many more can help you refine your image to get closer to what you want. For example: "Draw a ginger-and-white striped cat looking excited as it chases a mouse around a kitchen, in the style of an impressionist painter, with light streaming through the windows and prominent use of blue and yellow."
Add your preferred output
Also consider the format in which you want your image to be outputted, including orientation, aspect ratio, etc. or a particular use case, e.g. poster, email header image, social media post, etc.
Firefly also includes a sidebar that enables you to toggle many of these elements on and off, including effects, style, lighting, composition, and more. Learn how to get started on Firefly, or find more tips on creating effective prompts from Adobe.