AI Image & Video Tools
From Midjourney to Runway, AI can now generate stunning images and videos from text. Here's how to use these tools for real work — and avoid the pitfalls.
The pitch deck that almost didn't happen
Marcus runs a small design agency. A startup founder calls at 3 PM on Thursday: "We're pitching investors tomorrow at 10 AM. We need 12 custom illustrations for our pitch deck — futuristic healthcare scenes, product mockups, team culture photos. Our budget is $500."
A year ago, Marcus would have said no. Custom illustrations take 2-3 days per piece. Stock photos would look generic. There was no way to deliver 12 unique visuals in 19 hours for $500.
Instead, Marcus opened Midjourney. By 9 PM he had 30 candidate images — futuristic hospital lobbies, doctors using AR headsets, abstract data visualizations. He refined 12 of them, composited a few in Canva, and delivered the deck at 8 AM. The founder closed a $2M seed round.
Marcus didn't replace his design skills. He used AI as a starting point — a way to go from blank canvas to first draft in minutes instead of hours.
By the end of this module, you'll know the four major AI image tools and when to use each, write prompts using the six-part image prompt formula, and understand where AI video generation is heading.
The AI image generation landscape
Four tools dominate AI image generation, each with distinct strengths.
| Tool | Best for | Style | Access | Cost |
|---|---|---|---|---|
| Midjourney | Artistic, stylized imagery | Painterly, cinematic, polished | Discord bot or web app | $10-$60/mo |
| DALL-E 3 | Quick concept images | Clean, illustrative, literal | Built into ChatGPT | Included with ChatGPT Plus ($20/mo) |
| Stable Diffusion | Full control, customization | Anything (open-source) | Local install or web UIs | Free (self-hosted) or $10/mo (DreamStudio) |
| Adobe Firefly | Commercial-safe images | Professional, stock-photo feel | Adobe Creative Cloud | Included with CC ($23/mo) or free tier |
✗ Midjourney
- ✗Stunning artistic quality
- ✗Requires Discord or separate web app
- ✗Better for mood, atmosphere, style
- ✗Strong with abstract and creative prompts
- ✗More steps to get started
✓ DALL-E 3
- ✓Built into ChatGPT — no separate tool
- ✓Better at following literal instructions
- ✓Stronger with text in images
- ✓Easy to iterate in conversation
- ✓Good enough for most business uses
Writing prompts that actually work
The gap between a mediocre AI image and a stunning one is almost always the prompt — the same principle from the Claude module (specificity beats vagueness), but applied to visual output. Here's the formula that works across all tools.
The anatomy of a good image prompt:
[Subject] + [Style/Medium] + [Lighting] + [Composition] + [Mood] + [Details]
| Component | Example | Why it matters |
|---|---|---|
| Subject | "A woman reviewing data on a holographic display" | What's in the image |
| Style | "Digital illustration, clean vector style" | Determines the visual feel |
| Lighting | "Soft blue ambient lighting" | Sets mood dramatically |
| Composition | "Wide shot, rule of thirds" | Controls framing |
| Mood | "Professional, futuristic, optimistic" | Emotional tone |
| Details | "Minimal UI elements, glass desk, dark background" | Specificity matters |
Weak prompt: "A business meeting"
Strong prompt: "Professional team of four people collaborating around a glass conference table, modern office with floor-to-ceiling windows, natural daylight, candid photography style, diverse team, laptops and notebooks on the table, shallow depth of field, warm and productive mood"
There Are No Dumb Questions
"Can I use AI-generated images commercially?"
It depends on the tool. Midjourney's paid plans grant commercial usage rights. DALL-E 3 via ChatGPT Plus allows commercial use. Adobe Firefly is specifically trained on licensed content to be commercially safe. Stable Diffusion depends on the model and license. Always check the specific tool's current terms of service before using images in commercial projects.
"Will AI replace graphic designers?"
No — but it's changing the job. Designers who use AI generate concepts faster, iterate more quickly, and spend more time on high-level creative direction. AI handles the first draft; humans handle taste, brand consistency, and final polish. The designers struggling are those who refuse to learn the tools.
AI for video generation
Video is where AI tools are evolving fastest. As of mid-2025, you can generate short clips from text or images.
| Tool | What it does | Best for | Cost |
|---|---|---|---|
| Runway Gen-3 | Text-to-video, image-to-video, video editing | Short cinematic clips, product demos | $12-$76/mo |
| Pika | Text-to-video with style control | Social media clips, stylized animations | Free tier / $8/mo |
| Kling AI | High-quality video generation | Realistic motion, longer clips | Free tier / varies |
| Sora (OpenAI) | Photorealistic video from text | (Availability varies — check openai.com for current access) | TBD |
Practical video workflows
Social media teasers — Generate a 5-second loop from a product image for Instagram Reels
Concept visualization — Show a client what a final video might look like before hiring a production team
Background footage — Create abstract or atmospheric B-roll for presentations
Ad prototyping — Test 10 different visual concepts before committing budget to a real shoot
Image editing with AI
Beyond generating images from scratch, AI tools now handle editing tasks that used to require Photoshop expertise.
| Task | Tool | How it works |
|---|---|---|
| Remove backgrounds | Remove.bg, Canva AI | One click — AI detects subject and removes background |
| Extend images | DALL-E outpainting, Photoshop Generative Fill | AI generates content beyond the original frame |
| Remove objects | Photoshop Generative Fill, Cleanup.pictures | Select an object, AI fills the space naturally |
| Upscale | Topaz Gigapixel, Magnific AI | AI adds detail to low-resolution images |
| Style transfer | Midjourney /describe + /imagine | Upload a reference image, generate new ones in that style |
Plan a real visual project
50 XPPick a real project you're working on (or will work on) and plan an AI-assisted visual workflow: 1. **The project:** What visuals do you need? (e.g., pitch deck, social campaign, website redesign) 2. **Tool selection:** Which AI image tool would you use for each visual, and why? 3. **Prompt drafts:** Write 2 specific prompts for images you'd need 4. **Editing plan:** Which AI editing tools would you use for refinement? 5. **Human touch:** What would still need a human designer's input?
Sign in to earn XPThere Are No Dumb Questions
"How do I get consistent style across multiple AI images?"
Use a style reference in every prompt. In Midjourney, use the --sref flag or upload a reference image. In DALL-E 3, describe the style explicitly in each prompt ("digital illustration, flat design, muted earth tones, thin line art"). In Stable Diffusion, use a consistent model checkpoint and style LoRA. Consistency is the hardest part of AI image work.
"Are AI images always obvious?"
Not anymore. The best AI images are nearly indistinguishable from professional photography or illustration. But telltale signs remain: text in images is often garbled, hands can look wrong, and backgrounds may have impossible geometry. Always review carefully.
When NOT to use AI images
✗ Use AI images
- ✗Internal presentations and mockups
- ✗Social media content that refreshes weekly
- ✗Blog post illustrations
- ✗Concept exploration and brainstorming
- ✗Rapid prototyping
✓ Use traditional methods
- ✓Brand photography that must feel authentic
- ✓Legal or compliance-regulated materials
- ✓Images where specific real people must appear
- ✓Final production assets for major campaigns
- ✓Situations where AI-generated content could erode trust
Back to Marcus's 3 PM phone call
Remember Marcus and his impossible deadline — 12 custom illustrations in 19 hours for $500? The reason he delivered wasn't just "he used Midjourney." It was that he knew the six-part prompt formula, generated 30 candidates (not 12), and refined the best ones in Canva. He used AI for the first 80% — concept, composition, mood — and applied his design judgment for the last 20%. The startup founder closed a $2M seed round. Marcus landed a retainer client. The prompt formula was the lever that made it possible.
Key takeaways
- Four major AI image tools: Midjourney (artistic), DALL-E 3 (convenient), Stable Diffusion (customizable), Adobe Firefly (commercially safe)
- Prompt quality determines output quality — use the formula: subject + style + lighting + composition + mood + details
- AI video generation is real but early — useful for short clips, prototyping, and social media, not for production video
- AI image editing (background removal, outpainting, upscaling) is mature and production-ready
- Always check commercial usage rights for your specific tool and plan
- Best workflow: AI for the first 80%, human refinement for the last 20%
Next up: AI isn't just generating images — it's writing code. The next module covers GitHub Copilot and coding assistants that can generate entire functions from a comment.
Knowledge Check
1.Which AI image tool is specifically designed to be commercially safe, trained on licensed content?
2.What is the most important factor in getting high-quality AI-generated images?
3.What is the current state of AI video generation tools like Runway and Pika?
4.What is the recommended professional workflow for using AI image tools?
Want to go deeper?
🧠 AI & Machine Learning Master Class
Understand AI, use it in your job, and build AI-powered products.
View the full program