Scenario 4 · Can I run this?
Upload a video creative to AI and receive four layers of analysis: quick verdict → shot-by-shot breakdown → platform deployment → new creative directions. This guide walks through a real 30-second air conditioner ad end-to-end.
Upload a video creative to AI and receive four layers of analysis: Quick Verdict → Shot-by-Shot Breakdown → Platform Deployment → New Creative Directions. This guide walks through a real 30-second air conditioner ad end-to-end — from "Can I run this?" to "How to fix it."
Who This Tutorial Is For
- You have a video creative on hand and aren't sure if it's ready to run
- Your ad has been live for a few days with weak metrics, and you want to know exactly what's wrong
- You want AI to give you specific "change this at 0:17" fixes and full new direction scripts — not just "consider optimizing"
What You Need
One video file. That's it.
Supported formats: MP4, MOV (just upload). You don't need to provide any product briefs or background info upfront — the video itself is the entire input.
Full Conversation Walkthrough
Step 1 · Upload Your Video + One-Line Prompt
You
↑ The creative being analyzed: a 30-second vertical portable AC ad
Analyze this video creative for me
GrowthGPT
GrowthGPT automatically handles the following (no extra input needed from you):
- Identifies video duration, aspect ratio, language, and target market
- Infers product category and core selling points
- Enters the four-layer analysis workflow
Step 2 · L1 Quick Verdict — A Decision in 30 Seconds
AI returns a five-dimension scorecard, each dimension scored 1–5:
| Dimension | Score | Rationale |
|---|---|---|
| Hook Strength (first 3s) | 3/5 | Opens with a relatable heat-discomfort scene, but operates at a conscious level only — no instinctive pattern-break to stop cold-scrolling users |
| Value Prop Delivery | 4/5 | Three selling points (fast cooling, easy setup, aesthetic fit) demonstrated clearly; "25°C" screen provides tangible proof |
| Visual Execution | 4/5 | Clean quality, consistent lighting, smooth cuts; shot density within optimal range for 30s |
| Platform Fit | 3.5/5 | Vertical 9:16, 30s — meets format specs; but pacing feels more "brand film" than platform-native content |
| CTA Effectiveness | 2.5/5 ⚠️ | Ends with a brand tagline, not an action command; no price, no urgency, no "Shop Now" |
Overall Verdict: ⚠️ Needs Revision | Average: 3.4/5
Key Insight: The production quality is solid, but the final seconds completely lack a "last push" — viewers finish feeling positive but have zero impulse to act. The CTA gap is the critical fault. A spelling error at 0:17 ("livingoms") is also flagged as a trust-damaging hard fault.

↑ L1 five-dimension scorecard · actual GrowthGPT output
Step 3 · L2 Shot-by-Shot Breakdown — See What Every Second Is Doing
You
Proceed to L2, give me a shot-by-shot structural breakdown
AI maps all 17 shots across the 30-second timeline, tagging each with its narrative role:
Structural Map
| Segment | Timecode | Share | Content |
|---|---|---|---|
| HOOK | 0:00–0:02 | 7% | Woman distressed by heat — problem establishment |
| DEMO + PROOF | 0:02–0:22 | 67% | Product reveal → setup demo → activation → airflow proof → multi-room scenes → interface close-ups |
| CTA (attempted) | 0:22–0:30 | 26% | Lifestyle recap + brand tagline — no actual call-to-action |
Pacing Curve

↑ Pacing curve — energy peaks early (~0:09) and never builds to a conversion climax
Key Insight: The emotional peak lands too early (~0:09, the "hair blowing in cool air" relief moment), then the video enters a flat plateau of supplementary info. The final third decreases energy rather than building to a conversion climax. Approximately 6 seconds (20% of airtime) are redundant — repeating proof points already established.
Narrative Structure
| Standard Formula | This Video | Status |
|---|---|---|
| Problem (hook the pain) | ✅ Shot 1 | Present but weak |
| Solution Introduction | ✅ Shots 2–6 | Strong |
| Proof / Demonstration | ✅ Shots 7–14 | Strong |
| Social Proof / Authority | ❌ Missing | Gap |
| Offer + CTA | ❌ Missing | Critical gap |

↑ L2 shot-by-shot breakdown · actual GrowthGPT output
Step 4 · L3 Platform Deployment — Where Should This Ad Run?
You
Proceed to L3
AI evaluates the video against each platform's specs, audience behavior, and content norms:
| Platform | Verdict | Reason | Adjustment Needed |
|---|---|---|---|
| Meta Reels | ✅ Ready (after fixes) | 9:16 native, 30s sweet spot, polished aesthetic matches ecosystem | Fix typo + add 3s CTA end card |
| Meta Feed | ⚠️ Needs adaptation | Strong demo content, but 9:16 crops to 4:5 in feed; no captions for sound-off viewers | Reframe to 4:5 safe zone + add subtitles + CTA |
| TikTok Spark Ads | ⚠️ Needs restyling | Vertical + product-focused, but too polished — feels like an ad, not native content | Re-edit with faster hook, kinetic text, casual VO |
| TikTok Shop | ⚠️ Significant rework | Missing: price, cart CTA, urgency badge, creator-style delivery | Add price overlay + "Tap the cart" + limited-stock badge |
| Google / YouTube | ❌ Poor fit | 9:16 incompatible with 16:9 requirement; no brand bumper in first 5s | Not worth adapting — produce a dedicated horizontal asset |
A/B Test Recommendations
AI also designs specific tests with hypotheses:
| Test | Version A | Version B | Key Metric |
|---|---|---|---|
| Hook | Current "woman fanning" | Pattern-break: ice cube melting in 3s time-lapse | 3s view rate |
| CTA | Brand tagline close | Price + countdown + "Shop Now" end card | CTR + Purchase CVR |
| Talent | Female protagonist | Male protagonist (same setting) | CPA by gender |
| Platform Style | Current polished edit | TikTok-native re-edit (kinetic text, trending audio) | CPM + CPA on TikTok |

↑ L3 platform deployment matrix · actual GrowthGPT output (excerpt)
Step 5 · L4 New Creative Directions — AI Writes Your Next Script
You
Show me several new creative directions
Based on all three prior layers, AI produces full creative direction briefs with scene descriptions, production notes, and validation metrics:
P1 "Temperature Race" — Lowest cost, highest learning value
Hook type: Curiosity gap — "I tested if this $199 AC can cool my apartment from 35°C to 25°C in under 10 minutes"
Structure: Creator talking-head challenge → time-lapse setup (stopwatch running) → temperature dropping in real-time → over-delivery surprise ("8 minutes — I'm actually freezing") → CTA with price
Why it's strong: The race format creates natural suspense. The thermometer provides second-by-second proof no polished brand video can replicate. Creator format matches TikTok/Reels native language exactly.
Production: 1 day, $300–$1,000. The shoot IS the content.
P2 "The 3AM Wake-Up" — Strongest instinct-level hook
Hook type: Survival threat — sweaty face in darkness, phone showing 3:17 AM, drenched pillow outline
Structure: Threat (3s, suffocating heat at night) → Hard cut to same person sleeping UNDER A BLANKET in morning light with AC running → Quick setup demo → Temperature proof → CTA with price + urgency
Why it's strong: Suffocation/trapped heat triggers a Level 1 biological response. The blanket-in-summer release passes the surprise test — no one expects that level of comfort from a portable unit.
Production: 2–3 days, $1,500–$3,000.
P3 "The Landlord Can't Say No" — Best for renters, high shareability
Hook type: Social threat — text overlay slams: "My landlord said NO to AC installation"
Structure: Social powerlessness (2s) → Rebellious smirk + product reveal on wheels (2s) → POV setup demo "10 seconds, done, landlord will never know" → Comfort proof (blanket + grin) → CTA: "Renters' secret weapon"
Why it's strong: Names the #1 real purchase blocker for renters. The power-shift moment (finding a no-permission-needed workaround) triggers autonomy/control emotion — one of the strongest purchase drivers for 25–35 renters.
Production: 1 day, $500–$1,500.
Bonus: P4 "Split Screen — Two Realities" — Best for retargeting
Hook type: Pattern-break — same room split in half; left side orange/hot/miserable, right side blue/cool/productive
Structure: Impossible visual contradiction (2s) → Cool side expands to full frame (product reveal) → Multi-room montage → CTA with price
Production: 1 day shoot + 2 days post, $2,000–$4,000.

↑ L4 new creative directions · actual GrowthGPT output
How Many Messages Did This Take?
One to start. Three follow-ups to go deeper.
| Your message | What you got back |
|---|---|
| "Analyze this video creative for me" | L1: Full five-dimension scorecard + verdict + issues + fixes |
| "Proceed to L2" | Shot-by-shot map of all 17 shots + pacing curve + structural diagnosis |
| "Proceed to L3" | Platform-by-platform deployment matrix + 4 A/B test designs + launch sequencing plan |
| "Show me new creative directions" | 4 production-ready creative briefs with scripts, hooks, budgets, and validation metrics |
AI automatically handled:
- Video specs identification (duration / aspect ratio / audio type)
- Product feature extraction (appearance / setup / controls / accessories)
- Decision model construction (who buys this, why, what stops them)
- Five-dimension scoring + hard-fault detection
- 17-shot structural analysis + pacing assessment
- 6-platform fit evaluation + specific adjustments per platform
- 4 new creative directions with full scene-by-scene scripts
From upload to complete four-layer analysis: approximately 2 minutes per layer.
What You Can Do Next
| Action | Related Tutorial |
|---|---|
| Fix the typo + add a CTA end card, then launch on Meta Reels | Launch a Campaign |
| Use the P3 direction to re-cut existing footage + generate a new thumbnail | Generate Ad Creatives |
| Feed the diagnosis into Autopilot to set creative rotation rules | Autopilot · Get Started in 3 Steps |
FAQ
Q: Does it work with horizontal video?
Yes. 16:9 / 1:1 / 9:16 are all supported. AI automatically adjusts platform-fit recommendations based on aspect ratio.
Q: Is there a length limit?
Videos under 60 seconds work best. Longer videos are still analyzed, but AI will recommend cutting shorter versions for paid media.
Q: Can I analyze multiple videos side-by-side?
Yes. Upload several and say "Compare these — which one should I run on TikTok?" AI will produce a head-to-head scoring comparison.
Q: Can I export the analysis?
You can copy the full output as Markdown. PDF export is coming in a future update.
Q: What if I already know the target platform?
Mention it upfront (e.g., "Analyze this for TikTok Shop") and AI will weight platform-specific criteria more heavily in the L1 scoring and skip the platforms that don't apply in L3.
Q: Can it analyze competitor video ads too?
Yes. Say "Break down this competitor ad" or "Analyze this — it's a competitor's creative" and AI switches to competitor teardown mode: highlighting what works, what's learnable, and how you can do it better.
Scenario 3 · Make my visuals
No designer? Let AI make the images. One description + one product photo, and get 4 ad creative variants in 5 minutes.
Scenario 5 · Spin up a new campaign
Want to launch a new ad? Give AI a product URL + your campaign goal, and get a ready-to-ship campaign back in 5 minutes.