Skip to main content

​
The Problem

Product pages with video significantly outperform those without — higher engagement, longer time on page, and better conversion rates. But with thousands of SKUs, creating individual product videos is impossible with traditional production. Most e-commerce stores have great product images but zero product video.

​
How It Works

Product catalog (name, description, images) → Prompt per product → Batch generate → Embed on product pages
Your product database already has everything Video Agent needs: name, description, features, and images. Turn that structured data into video at scale.

​
Build It

1

Pull product data

# From your catalog API, database, or CSV
products = [
    {
        "name": "CloudWalk Pro Running Shoes",
        "category": "Footwear",
        "price": "$129",
        "description": "Lightweight performance running shoe with responsive foam midsole and breathable knit upper.",
        "features": [
            "ResponsiveFoam midsole — 30% more energy return",
            "Breathable knit upper — keeps feet cool on long runs",
            "Carbon fiber plate — propels you forward",
            "Only 7.2 oz — one of the lightest in its class",
        ],
        "images": [
            "https://cdn.store.com/products/cloudwalk-hero.jpg",
            "https://cdn.store.com/products/cloudwalk-side.jpg",
            "https://cdn.store.com/products/cloudwalk-sole.jpg",
        ],
    },
    # ... hundreds more
]
2

Build category-aware prompts

Consider using different video styles for different product categories. For example, fashion might benefit from energy and aspiration, while electronics might call for clarity and specs.
CATEGORY_STYLES = {
    "Footwear": {
        "tone": "energetic and aspirational, like a Nike ad",
        "focus": "performance benefits, how it feels, lifestyle context",
        "duration": "20 seconds",
    },
    "Electronics": {
        "tone": "clear, knowledgeable, like a trusted tech reviewer",
        "focus": "specs that matter, real-world use cases, comparisons",
        "duration": "30 seconds",
    },
    "Home & Kitchen": {
        "tone": "warm, practical, like a friend recommending a product",
        "focus": "solving everyday problems, quality materials, ease of use",
        "duration": "25 seconds",
    },
}

def build_product_prompt(product):
    style = CATEGORY_STYLES.get(product["category"], CATEGORY_STYLES["Electronics"])
    features = "\n".join(f"- {f}" for f in product["features"])

    return f"""Create a {style['duration']} product video for {product['name']}.

Product: {product['name']} — {product['price']}
{product['description']}

Key features:
{features}

Video structure:
- Hook (3s): Bold statement about the key benefit
- Features (70% of duration): Walk through 2-3 standout features
  with text overlays. Reference the attached product images.
- CTA (3s): "Available now for {product['price']}"

Tone: {style['tone']}
Focus on: {style['focus']}
Use the attached product images as visual reference.
"""
3

Batch generate with rate limiting

import requests
import time

video_jobs = []

for product in products:
    prompt = build_product_prompt(product)
    files = [{"type": "url", "url": img} for img in product["images"][:5]]

    resp = requests.post(
        "https://api.heygen.com/v3/video-agents",
        headers={
            "X-Api-Key": HEYGEN_API_KEY,
            "Content-Type": "application/json",
        },
        json={"prompt": prompt, "files": files},
    )
    video_jobs.append({
        "product_id": product["name"],
        "video_id": resp.json()["data"]["video_id"],
    })
    time.sleep(5)

print(f"Submitted {len(video_jobs)} product videos")
Then poll for completion and match video URLs back to product IDs.
4

Embed on product pages

Once rendered, add video URLs to your product data and display on your store.
# After polling all videos to completion:
for job in video_jobs:
    update_product_page(
        product_id=job["product_id"],
        video_url=job["video_url"],
        thumbnail_url=job["thumbnail_url"],
    )

​
Video Types by Use Case

Video typeDurationWhen to use
Product showcase15–30sProduct listing page — show key features
How-to-use30–60sComplex products — demonstrate usage
Comparison30–45s”Pro vs Standard” — help buyers choose
Unboxing/first look20–30sNew arrivals — build excitement
Customer testimonial30sSocial proof — pair with review text

​
Scaling to Thousands of Products

For large catalogs, consider prioritizing by business impact:
  1. Top sellers first — highest traffic pages get the most ROI from video
  2. High-margin products — the conversion lift matters most here
  3. New arrivals — video helps customers understand unfamiliar products
  4. Products with high return rates — better video = more informed purchases = fewer returns
# Prioritize by revenue impact
products.sort(key=lambda p: p["monthly_revenue"], reverse=True)
top_products = products[:100]  # Start with top 100

​
A/B Testing

Generate multiple video styles for the same product and measure which converts better: 
variants = [
    {"style": "presenter explaining features", "label": "explainer"},
    {"style": "fast-paced montage with text overlays only", "label": "montage"},
    {"style": "customer testimonial style, first-person", "label": "testimonial"},
]

​
Variations

  • Seasonal campaigns: Regenerate product videos with holiday-themed prompts
  • Multi-language: Translate for international storefronts using Video Translation
  • Social ads: Generate portrait (9:16) versions for social media advertising
  • Bundle videos: Combine multiple products into “complete the look” or “bundle” showcase videos

​
Next Steps

Real Estate Listings

Same catalog-to-video pattern, applied to properties.

Social Media Pipeline

Distribute product videos across social platforms.