The Stock Photo Problem
You need an image of a nurse reviewing patient charts at a hospital workstation.
Stock photo search returns: generic white coat, staged smile, obviously posed. It doesn't match your content. But custom illustration costs $75-150.
So you use the stock photo. Learners can tell it's fake.
💡 The shift
AI generates images for $0.04-0.80 each. Specific to your content. Not perfect, but custom.
The Tools
| Tool | Cost | Best for | Output |
|---|---|---|---|
| DALL-E 3 (via ChatGPT Plus) | $20/month | Quick iterations, realistic scenes | 2 images per prompt |
| Midjourney | $10-30/month | Artistic styles, illustrations | 4 variations per prompt |
| Stable Diffusion | Free (self-hosted) or $10-30 (hosted) | High volume, full control | Unlimited |
For most IDs: Start with DALL-E 3. Easiest to use. Most reliable results.
The Basic Prompt
Start simple:
Example:
This works 70% of the time. When it doesn't, you refine.
The Better Prompt
Add details for control:
Example:
More details = more control.
What AI Does Well vs. What It Gets Wrong
✓ AI Strengths
- People in workplace settings
- Objects and products (for product training)
- Conceptual illustrations (abstract concepts visualized)
- Scenes and environments
- Icons and simple graphics
✗ AI Limitations
- Text within images (letters come out garbled)
- Complex diagrams with labels
- Precise technical schematics
- Specific real people (can't recreate your CEO)
- Brand-specific items (copyrighted logos, products)
💡 The Division of Labor
AI creates custom visuals fast. You handle text, diagrams, and brand assets with traditional tools—or hire a designer.
The Cost Math
| Method | Cost | Time | Control |
|---|---|---|---|
| Stock photo | $10-30/image | 15-30 min searching | Low (limited to what exists) |
| Custom illustration | $75-150/image | 1-3 days | High (exactly what you want) |
| AI-generated | $0.04-0.80/image | 2-5 min per attempt | Medium (iterative refinement) |
For routine e-learning visuals, AI wins on speed and cost.
The Refinement Loop
You won't nail it first try. Use follow-up prompts:
Before and After
❌ Stock photo approach
- Searched "customer service training"
- Found: staged photo, fake smile, headset, generic office
- Doesn't match retail context
- Cost: $15
✅ AI-generated approach
"Retail employee helping elderly customer at store checkout, warm interaction, modern retail environment, professional but friendly photography style"
- Specific to retail context
- Authentic-looking interaction
- Modern store visible
- Cost: $0.40
- Time: 3 minutes
Quality Check
Before using an AI image in your course:
| Check | Question |
|---|---|
| Clear subject | Is the main subject recognizable? |
| No distortions | Hands, faces, text all look normal? |
| Right tone | Matches the feel of your course? |
| Appropriate | Professional, inclusive, non-offensive? |
| Relevant | Does it actually support the learning content? |
🚩 Red Flag
AI occasionally creates strange artifacts—extra fingers, warped text, impossible physics. Always check before publishing.
Key Takeaways
- AI beats stock photos on customization and cost. $0.40 vs. $15 per image.
- Specificity improves results. Add details about subject, setting, style, composition.
- Iterate in real-time. First attempt won't be perfect—refine with follow-ups.
- Know the limits. AI can't do text, complex diagrams, or specific real people.
Try It Now
🎯 Your task:
Pick a concept from your current project that needs a visual. Generate 3 variations with AI. Run each through the quality checklist. Select the best one.
The test: Does it look more relevant than a stock photo would?
📥 Download: Image generation prompts and quality checklist (PDF)
Ready-to-use prompt templates for workplace scenes, conceptual illustrations, and more.
Download PDF