Skip to Main Content

Using Generative Artificial Intelligence for Research

Writing Effective Text to Image Prompts

A picture is worth a thousand words but what do you do when you need words to describe a picture that you are visualizing? 

Below are the suggested steps for crafting an image prompt. You don't necessarily need to follow or use all the steps but experiment with the strategies and your own words to produce the images you want. Be sure to check with your instructor if they permit using AI-generated images in your assignment. 

Steps:

  1. State the medium of the image (e.g., illustration, photograph, or painting)
    an illustration

  2. Describe the content or subject. You may also include a short description of the action, state, and/or mood. 
    an illustration of a college student 

  3. Add details (e.g. surroundings, background, colour scheme)
    an illustration of a college student wearing a blue backpack and waiting on a subway platform 

Illustration of a college student wearing a blue backpack and waiting on a subway platform
This image is generated in Adobe Firefly using the last prompt mentioned.

Additional Steps for Consideration: 

  1. Define the art style or genre (e.g., realistic appearance, pop art, hi-lo) 
    a minimalistic illustration of a college student wearing a blue backpack and waiting on a subway platform 

  2. Identify the image composition (e.g., resolution, lighting style, aspect ratio, camera view)
    a minimalistic illustration of a college student wearing a blue backpack and waiting on a subway platform in a 1:1 aspect ratio

(Macready, 2023; Rebelo, 2023)

  • Prompts can be written in natural language. However, you may want to use phrases and keywords separated by commas to keep things simple. For example, this prompt - a minimalistic illustration of a college student wearing a blue backpack and waiting on a subway platform in a 1:1 aspect ratio - can be rewritten as: 

minimalistic illustration, college student, with a blue backpack, waiting on a subway platform, 1:1 aspect ratio

  • Try to be specific in your description.  For example, instead of stating "a landscape painting," you could try "forest painting" or "waterfall painting." Note that concrete language (e.g., a cloudy day) produces more predictable results while abstract language (e.g., a moody day) generates more varied results.
  • Keep your prompts short. Using between 3 to 5 descriptive elements is recommended if you are new to using AI image generators. 
  • Understand the parameters of each AI tool. For example, in Midjourney, you need to enter /imagine before your prompt. Also, --no handles negative prompts in Midjourney (e.g., --no trees excludes trees in images generated). 
  • Experiment with different words to see how they impact the images generated. 
  • Acknowledge the use of AI when you use an AI-generated image in your own work. 

(Macready, 2023; Rebelo, 2023)

 

Learn more about the ethical and legal implications of AI-generated images. For example, avoid uploading someone else's copyright image as a reference into an AI tool since it could be a copyright infringement. If you use a prompt that draws reference to the specific style or work of an artist, you could potentially be plagiarizing and infringing on their work (refer to the Copyright Guidelines on Images for more information). 

It takes time and practice to craft and experiment with text to image prompts that get you the best results. You could try using a tool like ChatGPT to describe the image you want and get ideas for a prompt to use in an image generator. There are also online resources to browse AI-generated images and corresponding prompts:

AI Image Generators for Research and Creativity

AI image generators can be helpful collaborative tools for supporting visual research and enhancing creativity. AI is not a new technology for artists and has been utilized in different art and design industries including animation, filmmaking, interior design, and sculptural art. Here are some of the ways that AI image generators can help artists, designers, and filmmakers: 

AI-generated images can help an artist brainstorm their ideas, see different combinations of multiple concepts, define the aesthetic direction of a work, and create initial drafts for further development.

Karen X. Cheng used DALL-E 2 to explore her unique concepts and new possibilities to create the first magazine cover generated by AI for Cosmopolitan (Liu, 2022). Midjourney helped architectural designer Manas Bhati bring his surreal idea to life - residential skyscrapers decked out with plants, trees, and algae that purify the air (Stewart, 2022).

Some artists use AI to explore new perspectives and styles that they might not have thought of and push themselves outside of their comfort zone. “AI can analyze vast amounts of data and generate unique concepts that might inspire an artist's next masterpiece. It can create unexpected connections between diverse concepts, fueling the artist's creative spark.” (Fortino, 2023).

Rashaad Newsome, a mixed-media artist, leverages AI to work with different mediums such as collage, photographs, sculpture, and film with the aim of decolonizing narratives and placing Black and queer people at the centre (Clarke, 2023).

AI can help artists make improvements to their artworks by analyzing elements such as colour palette, pattern, and texture. These tools can provide suggestions based on inputted prompts and can be utilized as part of an artist’s iterative creative process. Also, AI art generators such as Photoshop AI and Stable Diffusion can help artists fix mistakes and inconsistencies as well as fill in gaps in their work.

Welcome to Chechnya is a documentary on the Russian government’s persecution of LGBTQ+ people. AI was used to recreate the voices of the late Anthony Bourdain and Andy Warhol to bridge archival gaps and to create “digital veils” to protect some of the interviewees in the film (Horton, 2023).

Receiving feedback and suggestions is crucial to artists to improve their craft, gain different perspectives on their work, and develop growth and confidence in their artistic style and visions. However, honest and constructive criticism can be hard to come by for some artists (Collins, 2023). Generative AI tools may help provide objective feedback and helpful suggestions. For example, in ChatGPT, an artist can input their work and receive suggestions for improving composition, colour, and other aspects of the artwork (Artit, n.d.).

Hannah Epstein developed Critbot for anyone to upload their painting, photograph, or another still visual media to receive a review based on the level of criticism they select: kind, constructive or harsh (Collins, 2023).

AI generators can automate certain tasks such as image editing. These tools can also quickly produce different iterations of an artist’s base work. Designers can use AI tools to conduct research and develop customized solutions for their clients. For example, interior decorating designers may use AI to optimize their work, such as space planning, creating mock-ups and mood boards, and receiving recommendations on colour schemes and furniture (HOMMÉS Studio, 2023).

Photographer Antti Karppinen took photos of models to build a dataset in several AI tools so he could produce new images based on the models in different locations, clothing, poses, etc. (Growcoot, 2022).

Learn more:

Connect with us