Adobe Firefly is a text-image generator launched in 2023. Since UNC students are eligible to receive Adobe Creative Cloud for free, it is a great option for text-image generation purposes. Like NotebookLM (profiled first), Adobe Firefly allows the user to upload their own images as structural and stylistic reference points. Such features improve prompt generation as they allow the user to better align the scene/image they have in mind with a pre-existing photo or style. (Adobe also now builds Firefly AI into its other image-based tools, but this review is only covering its text-image generator.)
Adobe Firefly: for Image Generation
Adobe Firefly allows users to:
- Create photorealistic and artistic images based on generative prompts
- Upload photos for stylistic and structural references
- Generate new iterations of images for social media content, flyers, and more
Features
- Generates images based on prompts
- Creates images in specific artistic styles
- Aligns images with structural and stylistic reference points
How to Use Adobe Firefly
- A student is writing an online popular science article on the dangers of PFAS in our daily lives and needs to include images to enhance their project. They have a specific topic and are struggling to find an adequate image from Creative Commons sites such as Unsplash. The student generates the following prompt: “Create a photorealistic image of a warning sign against PFAS.” Unfortunately, Adobe Firefly processes “PFAS” as “PAS” so the student has to revise the prompt so that the image generator creates an appropriate image
- A student wants to create a flyer for an event sponsored by the art club. They cannot find an adequate stock photo so they turn to Adobe Firefly to create an impressionistic drawing that better suits the event.
- Ask students to have Firefly AI generate images and have them explore its cultural biases in visual representation. For example, a student wants to generate an image of a scientist and asks Firefly AI to do so. What cultural biases does Firefly AI perpetuate in the images of scientists it generates? How can students overcome these biases by revising their prompts?
Watch out for. . . .
- If a key learning outcome is to generate accurate and precise images, then you might ask students to create a general text-image prompt before working with group members to refine the prompt to create a more accurate image
- If a key learning outcome is to understand how images can enhance and complement written work, you might ask students to describe five different appropriate images for a writing assignment before asking them to engineer text-image prompts to generate those images