Week 6 • Generative Imaging
To Do This Week
- Read: AI Co-creation: AI Images
- Bring: A digital photo of you has a child. Choose one that is in a location familiar to you. Make sure details are clear in the background. No headshots.
- Bring: Bring other digital photos that you shot and you like for aesthetic reasons, or digital copies of your own artwork or drawings. Something that has your visual sensibility.
Module Notes
Image Generators
-
DALL·E — Text-to-image generator known for following prompts closely and handling typography/scene layout well.
Technique: Diffusion (modern versions; earlier research used other approaches). -
Midjourney — Community-driven generator with strong aesthetics and stylization controls via Discord.
Technique: Proprietary diffusion. -
RunwayML — Creative suite for text-to-image and text-to-video (Gen-series models), background removal, and editing tools.
Technique: Diffusion. -
Stable Diffusion — Open-source text-to-image family (SD 1.x/2.x/XL) with many community checkpoints and control tools.
Technique: Latent diffusion. -
Vizcom — Turns sketches or design ideas into realistic renders, aimed at industrial designers and creatives. Upload a drawing or concept and get instant styled outputs.
Technique: Diffusion-based generative model (details proprietary). -
PromeAI — Design and rendering assistant (architecture, product, fashion) with style templates and AI upscaling.
Technique: Diffusion-based features (vendor specifics not fully disclosed).
In-Class ChatGPT Exercises
Work individually or in pairs. Use ChatGPT to build prompts, ask follow-up questions, and generate or refine images. Pay attention to how changes in wording, style, and parameters affect outputs.
- Hands: “a hand,” “a normal hand,” “clasped hands,” “holding hands,” “a hand with 4 fingers.” Examine results for race/gender bias or anatomical oddities. Ask why the model produced these alternate outcomes.
- Childhood Home or Room: Describe your childhood home or bedroom in detail (lighting, color, spatial position, textures). Iterate with ChatGPT prompts until you approach your memory. Experiment with aspect ratios and styles.
- Imaginary or Dream Image Describe an image from a dream or daydream in detail.
- Expanded Childhood: In Photoshop 2025, open a childhood photo (or a recreation). Use the Crop Tool to drag outward to add empty canvas where you want more image. In the contextual bar, choose Generative Expand. Optionally enter a prompt or leave it blank to auto-extend. Click Generate, review the 3 variations, and pick one. If edges look off, lasso the area and use Generative Fill to refine.
- Ekphrasis: Describe a vivid, dramatic fictional artwork—specify style, artist, materials, patterns, perspective. Add/remove details or change point of view.
- Upload an image and get a description of its content and style. Now have it generate an image based on the description.
- Abstraction: Prompt for abstract or conceptual images, e.g. “a non-representational image of…,” “a cubist image of…,” “a visualization of gravity,” or “a 4D representation.”
- Detail Expansion: Start with a small detail and keep elaborating step by step.
- Elaboration & Variation: Start with an adjective ("a fun image", "a bland image", "a boring image"). Ask for more/less: more fun, more boring, more bland. Or more shadow, less saturation, fewer people, smaller size, etc.
- Sketch to AI Image: Upload or create a simple sketch to Vizcom or similar tool and elaborate with prompts.
- Style Remix of Your Own Art: Upload a digital copy of your own art/sketch. With ChatGPT, draft 3–5 prompts that translate it into different art styles, movements or media, ie. comic style, anime style, film noir style.
- Extend Your Sketch into a Scene: Scan a quick pencil sketch and bring it into Photoshop. Use Generative Expand to widen the canvas into a full environment (street, forest, interior). Use Generative Fill to add supporting elements (floor, sky, props). Iterate with short prompts like “overcast light,” “late-afternoon shadows,” “wet pavement.”
- AI Collage from Your Assets: Cut out 3–5 elements from your own photos or artworks and arrange them on a new canvas. Ask ChatGPT for background/bridging prompts (e.g., “subtle paper texture with torn edges,” “foggy seaside horizon”). Use Generative Fill to synthesize a cohesive backdrop and soft shadows, then push 2–3 alternative moods (noir, pastel, high-contrast).
In-Class Series Activity
Develop a series of at 3-5 images based on these image experioments. Further develop a visual idea or technique. Share the images on Slack along with a brief description of process.
Themes to Explore
- Abstraction: Shapes, colors, forms to convey emotions or impressions.
- Surrealist Imagery: Dream-like scenarios that defy logic.
- New Species Creation: Imagine a new plant or animal species—environment, behaviors, appearance.
- Poetic Prompts: Translate the essence of a poem into imagery.
- Sketch to Image: Show evolution from simple sketch to detailed image.
- Photo-realism: Push detail and accuracy to create images resembling photographs.
- Comic Strip: Develop a comics panel of images based on some of your art.