Ideation in generative media is like panning for gold. You scoop a wide pan of river sand — most of it is worthless, but you're looking for those few glinting flakes that signal a vein worth mining. The goal isn't perfection; it's discovering directions you didn't know existed.
The Ideation Mindset
The biggest mistake in ideation is being too specific too early. You don't know what you want yet — that's the whole point. Ideation prompts should be open invitations, not detailed specifications. You're asking the model: "Show me possibilities," not "Execute my exact vision."
"A 35-year-old woman with auburn hair wearing a navy blue linen blazer, standing in a minimalist Scandinavian office with floor-to-ceiling windows, morning light from camera-left, shot at f/2.8 on Sony A7R V, editorial style"
"Professional woman, modern workspace, natural light, editorial feel" — then run 20 variations and see what the model produces across compositions, color palettes, and moods you hadn't considered.
Step-by-Step Ideation Workflow
Step 1: SEED PROMPTS (3-5 loose concepts) │ ▼ Step 2: BATCH GENERATE (20-30 per seed, fast mode) │ ▼ Step 3: FIRST PASS (spend 2 min scanning all results) │ Flag anything that catches your eye — don't analyze yet │ ▼ Step 4: CLUSTER (group flagged results by mood/direction) │ "These 4 feel warm and intimate" │ "These 3 feel dramatic and high-contrast" │ ▼ Step 5: REFINE SEEDS (write new prompts based on clusters) │ Use discovered moods to guide more specific prompts │ ▼ Step 6: SECOND BATCH (15-20 per refined seed, still fast mode) │ ▼ Step 7: SELECT FINALISTS (3-5 images to move into production)
Practical Settings for Ideation
Mood Board Construction
After ideation, your flagged results form a natural mood board. Group them visually: warm tones together, dramatic compositions together, minimal layouts together. This mood board becomes the creative brief for your production phase. It tells both you and the model (via reference images) what you're aiming for.
- Ideation is about discovering possibilities, not executing plans. Keep prompts loose and volume high.
- Use the cheapest, fastest model available — quality doesn't matter here, diversity does.
- Trust your gut reactions when scanning results. The images that catch your eye in a 2-second scan are usually the ones worth pursuing.
- Your ideation results become a mood board that guides the production phase.