5 Prompt Mistakes That Waste Your Sora 2 Credits
After analyzing 500+ failed Sora 2 generations, we found that 80% of failures come from 5 preventable prompt mistakes. Here's what to avoid—and what to do instead.
💸 The Cost of Bad Prompts
Each failed generation burns 1 credit on Plus ($0.20) or 2 minutes on Pro. These 5 mistakes are costing the average user $12-15/month in wasted credits.
Mistake #1: Overloading with Details
❌ Bad Prompt (73 words):
"A middle-aged woman with curly brown hair wearing a red blazer and black pants standing in a modern glass office with floor-to-ceiling windows overlooking a city skyline at sunset, holding a tablet in her left hand while gesturing with her right hand, smiling confidently at the camera, soft golden hour lighting, professional corporate environment, 4K quality..."
✅ Good Prompt (18 words):
"Professional woman presenting in modern office, city view background, golden hour lighting, confident gesture"
Why it fails: Sora 2 gets confused when processing 10+ conflicting details. It tries to honor all constraints and ends up generating nothing that looks right.
The fix: Focus on 3-4 core elements: Subject + Action + Setting + Mood. Let Sora fill in secondary details automatically.
Mistake #2: Vague Actions ("Walking," "Talking")
❌ Bad Prompt:
"Person walking down a street"
✅ Good Prompt:
"Person striding confidently down busy city street, camera tracking from side"
Why it fails: "Walking" and "talking" are Sora's weakest actions. They require complex motion physics (legs, lips) that the model struggles with. Results often have unnatural movement or frozen faces.
The fix: Use camera-focused actions instead of character actions:
- Instead of "person talking": "Person nodding while camera slowly zooms in"
- Instead of "person walking": "Camera panning across person mid-stride"
- Instead of "person dancing": "Silhouette moving rhythmically, backlit by neon lights"
Mistake #3: Requesting Text/Numbers in Scene
❌ Bad Prompt:
"Close-up of smartphone screen showing 'NEW MESSAGE' notification"
✅ Good Prompt:
"Close-up of smartphone screen glowing with notification, add text in post-production"
Why it fails: Sora 2 generates gibberish text 92% of the time. Even simple words like "OPEN" come out as "OPNE" or random characters.
The fix: Generate the scene without text, then add real text overlays in CapCut, Premiere Pro, or any video editor. This takes 30 seconds and looks 10x better.
⚠️ Exception: Background Signs
Blurry background text (like storefront signs 20 feet away) usually works fine. It's close-up legible text that fails.
Mistake #4: Ignoring Lighting Keywords
❌ Bad Prompt:
"Chef cooking in restaurant kitchen"
✅ Good Prompt:
"Chef cooking in restaurant kitchen, dramatic side lighting, steam illuminated by warm overhead glow"
Why it matters: Lighting is 40% of video quality. Without lighting keywords, Sora defaults to flat, uninspiring illumination that looks amateur.
Magic lighting keywords (add one to every prompt):
Outdoor Scenes:
- • Golden hour lighting
- • Soft diffused sunlight
- • Dramatic sunset backlight
- • Overcast natural light
Indoor Scenes:
- • Cinematic three-point lighting
- • Warm ambient glow
- • Dramatic side lighting
- • Soft window light
Mistake #5: No Camera Movement Specified
❌ Bad Prompt:
"Mountain landscape with lake reflection"
✅ Good Prompt:
"Mountain landscape with lake reflection, slow aerial drone shot rising from water level"
Why it fails: Static camera = boring video. Even gorgeous scenery looks lifeless without movement.
Essential camera movements (pick one per scene):
- Dolly in: "Camera slowly pushing toward subject"
- Tracking shot: "Camera gliding alongside subject"
- Crane up: "Camera rising from ground level to aerial view"
- Orbit: "Camera circling around subject" (Pro users only)
- Handheld: "Subtle camera shake, documentary style"
💡 Pro Tip: Dynamic Motion = Higher Retention
Videos with camera movement get 23% higher watch time than static shots (our analysis of 200 Sora 2 videos on YouTube).
Bonus: 5 Prompt Templates That Work 95% of the Time
Copy these proven templates and customize the [bracketed sections]:
1. Product Demo
"[Product] rotating on white background, studio lighting, camera orbiting 360 degrees, [color] accent lighting"
2. Nature Scene
"[Location] landscape, golden hour lighting, aerial drone shot rising slowly, [weather condition], cinematic color grading"
3. Person in Action
"[Person] [action] in [setting], camera tracking from [angle], [lighting style], shallow depth of field"
4. Abstract/Conceptual
"[Object/element] floating in [environment], slow motion, [color] lighting, camera slowly zooming in, ethereal atmosphere"
5. Urban/Cityscape
"[City location] at [time of day], [weather], camera [movement], [prominent element] in foreground, depth of field"
Your Action Plan
- Audit your last 5 prompts - Which of these 5 mistakes did you make?
- Rewrite 1 failed prompt using the fixes above
- Save successful prompts in a Google Doc for reuse
- Start with templates - Customize the 5 templates before writing from scratch
Master these fixes and your success rate will jump from 60% to 95%+. That's the difference between wasting $15/month and getting exceptional results every time.