The Problem
E-commerce product listing requires visual assets at scale. Each SKU needs: 5 white-background shots, 4 lifestyle scenes, 3 model photos, 2 marketing banners, and 1 showcase video.
Industry pricing (2026 China market reference): basic product photography runs $110-700/set, full shoots with models and video reach $1,400+. For 20 products, that's tens of thousands just on visuals.
As a developer, when I see repetitive work, I think about scripting. So I went looking for an automation-first solution.
Why GUI Tools Failed Me
Every AI image tool I tried only solved one step:
- Tool A does background removal, not model try-on
- Tool B generates great images but loses product details
- Tool C is full-featured but costs $30-70/month with image caps
The real problem wasn't quality — it was the manual pipeline between tools. Download from A, upload to B, download from B, upload to C. Repeat 20 times.
The Fix: CLI = Scriptable = Batchable
I discovered Bailian CLI (bl) — it wraps image generation, image editing, and video generation into a single command-line tool.
Setup
npm install -g bailian-cli
bl config set api_key YOUR_KEY
Get a free API key — starter credits cover this entire tutorial.
My 6-Step Pipeline
Product: Vintage brown leather crossbody bag.
Step 1: White-background product shot
bl image generate --prompt "vintage brown leather crossbody bag, white background, product photography, 45-degree angle"
For example: "Generate 4 e-commerce hero images in different scene styles, covering dining, kitchen, home, and café scenarios" — just describe what you need in plain language.
Step 2: Lifestyle scene
bl image edit --image ./product-white.png --prompt "Place bag on rustic café table, warm lighting, lifestyle feel"
For example: "Generate a suitable background for the product in the image, conveying a refreshing, cool sensation"
Step 3: Virtual model try-on
bl image edit --image ./product-white.png --reference ./model-photo.png --prompt "Model wearing this crossbody bag naturally, maintain pose"
For example: "Virtual try-on of clothing onto my own body, keeping my face, body shape, and pose unchanged"
Step 4: Style variants
bl image edit --image ./scene.png --prompt "6 style variants: warm soft, cool minimal, vintage film, Japanese fresh, dark luxury, vibrant Instagram"
For example: "Generate 6 different style variants: warm soft-light, cool minimal, vintage film, fresh Japanese, dark luxury, vibrant Instagram" — exact same wording works.
Step 5: Marketing banner
bl image generate --prompt "E-commerce banner, bag centered, bold '50% OFF', gold accents, dark background" --size 1280x720
For example: "Generate a high-quality e-commerce marketing poster suitable for Taobao and Xiaohongshu platforms"
Step 6: Product video
bl video generate --image ./scene.png --prompt "Product rotating on turntable, warm lighting, bokeh" --duration 5
For example: "Generate a product showcase video suitable for Douyin/TikTok and Kuaishou platforms"
Batch Mode
for product in crossbody-bag tote backpack wallet belt; do
bl image generate \
--prompt "${product}, white background, product photography" \
--output "./output/${product}-main.png"
done
20 SKUs done in one coffee break.
Cost Comparison
Based on Bailian official pricing (July 2026): qwen-image-2.0-pro = $0.071676/image, wan2.6-i2v 720P = $0.086012/sec.
| Traditional (China market) | CLI (official pricing) | |
|---|---|---|
| Total per SKU | ~$780-1,850 | ~$1.44 |
| Time per SKU | 5-7 days | 15-20 min |
| Batch capability | Linear | Scriptable |
Traditional pricing: 2026 industry survey data. CLI cost: 15 images × $0.072 + 5s video × $0.086.
Honest Take
Key limitations to note:
Works great for: Small-to-medium sellers, daily listings, flash sale prep
Limitations: Text rendering occasionally glitches, fine metal details (buckles/zippers) can "mutate", luxury goods still need real photography
Get Started
If you're the "script everything" type, this might be the highest-ROI tool discovery of the year for e-commerce visuals.
Top comments (0)