6. Generate AI art about your assigned fiction

• You must include images from at least three different image generators
• At least one of the images should be generated with absolutely no mention of the story or the names of the characters - just a description in words of the person or setting. (If you are using ChatGPT or Gemini for this, you will need to do it in a new chat, so it doesn’t have memory of what story you are talking about. See some examples here, where I used a separate window of ChatGPT to help me make the prompts, and this one to create the images. Notice that several of the prompts blended into each other, so you might want a new chat occasionally to avoid that.)
• At least one image should be abstract, trying to represent emotional states you felt while reading or watching the fiction, rather than a literal depiction of anything in the story.
• You should include several different artistic styles, not just the one that people associate with the story.

Experiment with several different image generators, and with a variety of different styles - maybe different historical painting styles, digital art, animation, camera angles, lenses, and lighting. If your assigned fiction was a movie or TV show, try some versions where you try to imitate the visual style of the original (with or without mentioning it in the prompt) and some versions where you try to make it create a very different visual style, or imagine a very different looking person in the role. If you have Frankenstein, see if you can find a description of the monster in the book, and get it to draw that, rather than the iconic version from the 1931 movie that everyone imagines nowadays.

Save these images in a folder on your computer so that you can use them when you are making your final presentation, and make a note of the prompts that you used to generate them.

Create a document that contains ten of the most interesting images (including the ones mentioned above), each with information about the model you used and the prompt that generated it. They might be interesting because the exact same prompt led to different results on different models, or because you think the image is beautiful, or because the model failed in some interesting way, or anything else. (You might need to add a couple sentences explaining what you found interesting about some of them.) Make sure that you include some that are very stylistically different from each other.

All of these have limits for free accounts, but you can keep making throwaway accounts.

• Within LLMs - these give you less direct control over the image prompt, but include a lot of general understanding of the world
  • ChatGPT (free account gives 2-3 images per day, paid account gives more usage)
  • Gemini (free account gives about 20 images per day, paid account gives more usage)
• Stable Diffusion - an open source model that gives you a lot of control of the image, but needs you to explain things in great detail (it doesn’t know history or stories or anything other than image captions)
  • https://dreamstudio.ai/ (free account needed - probably only one test generation per account) the official online site run by Stability.ai, the company that developed Stable Diffusion
  • If you have a powerful computer, you can go to https://stability.ai/ and install the full model on your computer to get unlimited generation - see this video for a setup guide
• Flux - another open source model that gives you a lot of control of the image, but needs you to explain things in great detail (it doesn’t know history or stories or anything other than image captions)
  • https://flux-ai.io/ official site (you have to select Flux Schnell or similar to get one that uses less than your 10 free credits per day)
• CrAIyon - a site with a free generator that I know less about, but seems to allow unlimited free generations without a login
• MidJourney - the most powerful tool that doesn’t require installing your own AI system, but requires a subscription ($10/month or more)
• Students in previous terms have also used systems with the following names - I haven’t checked if they still exist, if they are free, if they steal your information, or what their URL is:
  • Adobe Firefly
  • Brand Studio
  • Bria AI
  • Canva
  • Creative Fabrica AI
  • DaVinci
  • DeepAI
  • Flyne AI
  • Vidu
  • YeriAI

• Tutorial for OpenArt AI, which also has advice relevant for others
• Stable Diffusion Prompt Guide, again, many tips are relevant for all
• Prompt Engineering Tutorial: Text-to-Image - I found this guide the most helpful, but it’s 20 minutes long (the most helpful part begins at 4:30)

Assignments 1 2 3 4 5 6 7 8 9 · Final project

Module 6 Back to AI Literacy