Product Video Ad

What you're buying: a download package (skill instructions + Python scripts) that runs on your own computer inside your own AI agent — it is not a cloud service.
You need: an AI agent that supports skills (e.g. Claude Code or Codex) — its subscription or API costs are yours; Python 3.9+ and ffmpeg.
Third-party services: Microsoft edge-tts voice-over (free, internet needed). No image generation — you supply the product photos.

Product photos in, a ready-to-run short video ad out. No editing software, no AI image edits.

What you get

• 🎬 9:16 MP4 (TikTok / Reels / Shorts / Stories), plus optional 1:1 and 4:5 (feed ads)
• 🏷️ Exact price badge: shows your price character for character, with optional strikethrough original price and labels like "Launch deal" — it never invents a price
• ✨ Animated benefit captions: one point per scene, fade-in and slide-up
• 🗣️ Voice-over and subtitles: 400+ voices (English, Chinese, Japanese, Korean, Spanish…), subtitles wrap automatically
• 🎵 Background music: synthesized on the fly, so no copyright issues; or use your own licensed track
• 🔚 CTA end card: product image, price and buy button
• 📋 Ad copy: primary text, headline and hashtags

Example

A fictional insulated bottle, "ChillMate":

• English 9:16 ad, 15 s, with a "Launch deal $34.99 → $24.99" badge
• Traditional Chinese ad, 18 s, in 9:16 and 1:1, opening with the pain-point hook "水一下就變溫？" ("Water warm already?") and ending on "NT$590 · 點連結購買" ("NT$590 · Tap to buy")

The demo uses AI-generated sample photos of a made-up product. With your own photos, the ad shows your real product.

How it works

1. Provide 1–5 product photos, the product name, 2–4 benefits, the price and a CTA.
2. Your agent picks the best of 8 ad angles (pain → solution, before/after, 3 reasons, gifting, unboxing, comparison, social proof, price-led), writes the script and asks you to confirm.
3. The script renders the video: blurred background with the product hero shot, push/pull motion, captions, price badge, voice-over, subtitles, music and end card.
4. The agent samples frames to check that captions, price and subtitles don't cover the product, then hands over the video and copy.

Why no AI-generated product shots

AI edits make the product "look different", which leads to returns, complaints and possible breaches of ad-platform rules on misleading ads. This skill uses only your real photos.

Requirements

• Python 3.9+ and ffmpeg (the skill checks your setup and tells you what to install)
• pip install -U edge-tts pillow

Limitations (please read)

• Ad copy is AI-written — confirm every claim is true before running the ad.
• Voice-over is synthetic (Microsoft neural voices); label AI content where platforms require it.
• Quality depends on your photos: good lighting, 1000 px or larger and in-use shots work best.
• No automatic background removal; white-background or lifestyle photos are recommended.
• It does not place or upload ads and does not guarantee ad performance.
• Voice-over needs internet (Microsoft speech service); if the service changes, the author will keep it updated (download buyers get free updates for life).

Files

• SKILL.md — agent workflow
• scripts/make_ad.py — renderer
• references/ad-angles.md — 8 ad angles and hook formulas
• references/compliance.md — ad compliance checklist
• references/example_en.json / example_zh.json and sample_photos/ — ready-to-run examples