One-Photo Product Ad · Open Skill 公開技能
One-Photo Product Ad
Drop in one product photo and get a 10-second vertical ad built for TikTok, Reels and Shorts. What you get: - A 4-beat shot plan: hook (0-2s), reveal (2-5s), benefit (5-8s), CTA (8-10s) - Camera motion, lighting and background locked to one visual style so every frame sells the same product - Two caption tracks: on-screen hook text and a ready-to-record voiceover line - Square-safe framing, no warped logos, no invented claims How to use: 1. Upload one product photo (plain background works best). 2. Say what the product is and the one promise you want viewers to remember. 3. Pick a mood: clean studio / warm lifestyle / high-energy street. 4. Run it. You get the finished clip plus the shot list to tweak. Best for: DTC brands, e-commerce sellers, app promotion, launch day posts. Not for: multi-product catalogs, long explainers, or anything needing real people on camera.
SocialAd
3 0 1
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/4203025/2026-10-10/g2cpuxef.png
---
name: one-photo-product-ad
description: Turn one product photo into a 10-second vertical social ad with a locked 4-beat shot plan, consistent lighting, and ready-to-post captions. Use whenever the user uploads or points to a single product image and asks for an ad, promo clip, launch post, Reels/TikTok/Shorts spot, or product showcase. Also trigger on "make an ad from this photo", "10s product video", "turn this into a reel", "产品短视频", "电商主图视频", "给这个产品做个广告".
license: MIT
version: 1.0
---
# One-Photo Product Ad
Turn a single product photo into a 10-second vertical ad that looks like one continuous shoot, not four random clips.
## Input contract
**Required**
- One product image. Plain or near-plain background gives the cleanest results.
- The single promise the viewer should remember — one sentence, one benefit.
**Optional, ask once if missing**
- Mood: `clean studio` / `warm lifestyle` / `high-energy street` (default `clean studio`)
- Brand color, if any (default: derive from the product's dominant color)
- On-screen language (default: same language the user writes in)
**Refuse or hand back when**
- The image contains multiple unrelated products → ask which one is the hero, or suggest a catalog-style skill.
- The request needs a real person on camera → say so up front; this skill sells the object, not a spokesperson.
- The user asks for medical, financial, or before/after claims → strip the claim, keep the product.
## The 4-beat structure
Every ad is 10 seconds, four beats. Do not add a fifth beat — 10 seconds is a hard budget.
| Beat | Time | Job | Motion |
|------|------|-----|--------|
| 1. Hook | 0.0–2.0s | Stop the scroll before any context lands | Fast push-in or whip, product partly off-frame |
| 2. Reveal | 2.0–5.0s | Show the whole product, the "oh" moment | Slow orbit or glide, settles to hero angle |
| 3. Benefit | 5.0–8.0s | Show it in use or in hand, the "I want it" moment | Handheld-adjacent drift, shallow depth of field |
| 4. CTA | 8.0–10.0s | Product centered, logo readable, room for the button | Locked-off, gentle push, hold last 0.5s |
Why this order: the first two seconds are bought with motion, not information. Product identity lands in beat 2, desire in beat 3, and the final frame must be a still the viewer can read while the CTA button is on screen.
## Visual lock
Pick these once and reuse them in every beat prompt. Changing any of them between beats is what makes AI ads look stitched together.
- **Lens**: 50mm equivalent, f/2.0–2.8 for shallow depth unless the product is large
- **Light**: one key light at 45°, one soft fill, one rim for edge separation. Name the direction every time (e.g. "key from upper left")
- **Background**: one setting for all four beats. Beat 3 may add a contextual prop; the background itself does not change
- **Color**: product color stays true. Do not shift saturation between beats
- **Camera height**: keep the eye-line constant. Product ads break when the camera jumps from table level to overhead
Write each beat prompt in this shape:
```
Beat {n} — {job}
Subject: {exact product from the reference image, unchanged}
Camera: {lens}, {movement over time}, {start framing} → {end framing}
Light: {key direction}, {fill}, {rim}; {mood} contrast
Background: {locked setting}
Duration: {n} seconds, {fps}fps, 9:16
Avoid: text artifacts, warped logo, extra hands, background swap, color shift
```
## Captions and voiceover
**On-screen text** — maximum 6 words per card, no sentence-ending period.
- Beat 1: the hook as a question or a bold fragment
- Beat 3: the benefit, not the feature
- Beat 4: the CTA, three words max ("Shop the drop")
**Voiceover** — one line, under 2.5 seconds, written to be read aloud. Test it by reading it out; if it trips, rewrite it.
**Never** put text over the product's logo. Place captions in the lower third with a safe margin of 8% from the frame edge.
## Claim discipline
Ads get taken down for claims, not for visuals. Before delivering, chec